Structure and Performance of Cements, 2nd Edition

Structure and Performance of Cements

Structure and Performance of Cements Second edition

Edited by

J.Bensted and P.Barnes

London and New York

First published 2002 by Spon Press 11 New Fetter Lane, London EC4P 4EE Simultaneously published in the USA and Canada by Spon Press 29 West 35th Street, New York, NY 10001 Spon Press is an imprint of the Taylor & Francis Group This edition published in the Taylor & Francis e-Library, 2008. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” ©2002 Spon Press All rights reserved. No part of this book may be reprinted or reproduced or utilized in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. The publisher makes no representation, express or implied, with regard to the accuracy of the information contained in this book and cannot accept any legal responsibility or liability for any errors or omissions that may be made. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Structure and performance of cements/J.Bensted and P.Barnes [editors].—2nd ed. p. cm. Includes bibliographical references and index. 1. Cement. I. Bensted, J. II. Barnes, P. (John) TA434. S94 2001 624.1′833–dc21 ISBN 0-203-47778-2 Master e-book ISBN

ISBN 0-203-78602-5 (Adobe eReader Format) 0-419-23330-X (Print Edition)

2001020916

Contents List of contributors

xii

Preface

xvi

Notation 1

Cement manufacture Wieslaw Kurdowski 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 1.11 1.12 1.13

2

xviii

Introduction Raw material preparation Raw material crushing Raw material grinding Classifiers Homogenization Clinker burning Coolers Cement grinding Automation Environmental protection Acknowledgements References

1 1 2 2 3 4 5 7 13 16 18 20 23 23

Composition of cement phases Herbert Pöllman

25

2.1 2.2

25

2.3 2.4 2.5 2.6 2.7

2.8 2.9

Introduction Simple oxides, elements and their potential role in cement chemistry Calcium silicates Aluminium silicates, calcium aluminium silicates, their properties and occurrence in cements Calcium aluminates Calcium aluminium ferrites Influence of minor components (alone or in combinations)— alkalies, MgO, SO3, phosphates, fluorides, borates, chlorine, titanium, heavy metals Increased MgO content Sulphate-containing phases

27 27 34 35 38

41 41 41 v

Contents

2.10 2.11 2.12 2.13 2.14 2.15 2.16 3

Hydration of Portland cement E.M.Gartner, J.F.Young, D.A.Damidot and I.Jawed 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

4

5

vi

Chlorine-containing phases Phosphate in cement Heavy metals in cement Elements used as fluxes (boron and fluorine) Quantification of cementitious materials by Rietveld method Acknowledgements References

Introduction Hydration of tricalcium silicate Hydration of dicalcium silicate Hydration of tricalcium aluminate Hydration of the ferrite phase Hydration of Portland cement Microstructure of hardened paste Appendix—Glossary of terms References

41 44 46 46 50 51 51 57 57 57 78 81 87 89 100 108 108

Calcium aluminate cements John Bensted

114

4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10 4.11 4.12 4.13 4.14 4.15 4.16 4.17 4.18 4.19 4.20

114 115 116 117 118 123 124 126 126 127 127 130 133 134 134 136 136 137 138 138

Introduction Manufacture Phase composition Physical properties Hydration Admixtures Blends with other materials Low-temperature applications High-temperature applications Hydrophobic applications Durability Safe usage of CACs Concrete society report in the UK on CACs CAC and structures Further comments on sulphate resistance Thermal analysis methods for assessing CAC conversion Further comments Conclusion Acknowledgements References

Properties of concrete with mineral and chemical admixtures S.Chandra

140

5.1 5.2

140 141

Introduction Binders

Contents

5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 6

7

8

Interrelation of cement properties to concrete Hydration of cementitious materials Superplasticizers Interfacial transition zones (ITZ) Durability properties Engineering properties Need for development of long-term test methods Concluding remarks Acknowledgements References

142 143 150 156 163 168 178 179 180 180

Special cements A.K.Chatterjee

186

6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 6.11

186 186 188 189 206 218 224 226 231 232 233

Introduction Prior reviews and past definitions Drivers for special cement development Construction cements with enhanced durability potential Cement formulations with improved engineering properties Cementitious products for environmental benefits Simultaneous production of energy and cement Cements with high energy conservation potential Miscellaneous hydraulic and chemical cements Conclusion References

Developments with oilwell cements John Bensted

237

7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 7.11 7.12

237 238 239 239 240 241 242 244 246 251 251 251

Introduction ISO classes and grades Relevant ISO standards Brief comments on the ISO cementing standards Effects of different conditions Some other interesting phenomena with Class G and H cements Cementing in hostile environments Cements for some given well types Alternatives to standard Class G and H cements Conclusion Acknowledgements References

Gypsum in cements John Bensted

253

8.1 8.2 8.3 8.4

253 253 255 257

Introduction The calcium sulphate-water system The role of gypsum in set regulation of Portland cement-based systems Gypsum quality

vii

Contents

8.5 8.6 8.7 8.8 8.9 8.10 8.11 8.12

By-product gypsums Flash set False set Air set Portland cement-calcium aluminate cement compositions Calcium sulphoaluminate cements Conclusion References

9 Alkali-silica reaction in concrete D.W.Hobbs 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 9.10 9.11 9.12

Introduction The reaction Mechanism of expansion Concretes affected Visual and internal cracking induced by ASR Pessimum behaviour Sources of alkali Reactive silica Diagnosis of ASR as the cause of visual cracking Factors influencing expansion Concluding remarks References

10 Delayed ettringite formation C.Famy, K.L.Scrivener and H.F.W.Taylor 10.1 10.2 10.3 10.4 10.5 10.6 10.7 10.8 10.9 10.10 10.11 10.12

Introduction DEF in field concretes Coarse microstructure of materials damaged by DEF Macroscopic properties associated with DEF Effect of cement composition Cement hydration at 70–100°C Chemistry of changes after cooling to ambient temperatures Paste microstructure in materials cured at elevated temperature Expansion Conclusions Acknowledgements References

11 Chloride corrosion in cementitious system Wieslaw Kurdowski 11.1 11.2 11.3 11.4 11.5 viii

Introduction Chloride ions diffusion in cement paste Binding capacity of chloride ions in cement paste Factors influencing the diffusion of chloride ions in the cement paste Mechanism of cement paste destruction in the chloride medium

258 260 260 261 262 262 263 263 265 265 265 266 267 267 269 270 271 272 273 280 280 282 282 282 283 283 284 284 286 287 288 292 293 293 295 295 295 297 298 300

Contents

11.6 11.7

Chloride-induced corrosion of reinforcement in concrete References

12 Blastfurnace cements E.Lang 12.1 12.2 12.3 12.4 12.5 12.6

History Slag composition and reactivity Grindability Blastfurnace cement characterization Durability References

13 Properties and applications of natural pozzolanas F.Massazza 13.1 13.2 13.3 13.4 13.5 13.6

Introduction Classification of natural pozzolanas Pozzolana-lime mixes Pozzolana-containing cements Conclusions References

14 Pulverized fuel ash as a cement extender Karen Luke 14.1 14.2 14.3 14.4 14.5

Introduction PFA formation Characteristics of PFA Role of PFA on performance of extended systems References

15 Metakaolin as a pozzolanic addition to concrete Tom R.Jones 15.1 15.2 15.3 15.4 15.5 15.6 15.7 15.8 15.9 15.10

Introduction Structure of metakaolin (mk) Pozzolanic reactions of mk Effect of mk on the basic properties of PC concrete Effect of mk on the properties of uncured concrete Effect of mk on the properties of hardened concrete Durability of mk concrete Metakaolin in engineering concrete Acknowledgements References

16 Condensed silica fume as a cement extender H.Justnes 16.1 16.2 16.3

Introduction Physical effects Chemical effects

306 308 310 310 310 313 314 318 323 326 326 326 328 335 348 348 353 353 354 355 357 369 372 372 373 374 378 381 385 389 393 395 395 399 399 399 401 ix

Contents

16.4 16.5

Conclusions References

17 Cement-based composite micro-structures Stephen P.Bailey, David O’Connor, Sally L.Colston, Paul Barnes, Herbert Freimuth and Wolfgang Ehrfeld 17.1 17.2 17.3 17.4 17.5 17.6 17.7 17.8 17.9 17.10

407 407 409

Introduction Cement-composite development Micro-mould design Mould fabrication MSCCD applications MSCCD production Quality of MSCCD features Conclusions Acknowledgements References

409 411 412 412 413 415 416 419 419 419

18 X-ray powder diffraction analysis of cements J.C.Taylor, L.P.Aldridge, C.E.Matulis and I.Hinczak

420

18.1 18.2 18.3 18.4 18.5 18.6 18.7 18.8 18.9 18.10 18.11 18.12 18.13 18.14 18.15 18.16

Introduction X-ray diffraction from cement Qualitative analysis of cements The method of X-ray diffraction Physical factors affecting XRD quantitative analysis Choice of radiation in cement XRD studies Amorphous (non-diffracting) content Background intensity Rietveld refinement of OPC XRD patterns Refinement strategy for Rietveld XRD assays of cements Errors in Rietveld quantifications Crystal structure analysis of Portland cement phases by the Rietveld method Quantitative analysis of hydrating phases Recent Rietveld studies of real cements Conclusions References

19 Electrical monitoring methods in cement science W.J.McCarter, G.Starrs and T.M.Chrisp 19.1 19.2 19.3 19.4 19.5 19.6 x

Introduction Immittance formalisms Application of electrical measurements Concluding remarks Acknowledgements References

420 421 422 423 425 427 427 429 429 432 434 434 435 436 437 437 442 442 442 446 453 454 454

Contents

20 Nuclear magnetic resonance spectroscopy and magnetic resonance imaging of cements and cement-based materials Jørgen Skibsted, Christopher Hall and Hans J.Jakobsen 20.1 20.2 20.3 20.4 20.5 20.6 20.7

Introduction Solid-state NMR methods Structure and bonding in cement minerals Proton relaxation and pore structure Other nuclei Magnetic resonance imaging References

21 The use of synchrotron sources in the study of cement materials Paul Barnes, Sally L.Colston, A.C.Jupe, S.D.M.Jacques, M.Attfield, R.Pisula, S.Morgan, C.Hall, P.Livesey and S.Lunt 21.1 21.2 21.3 21.4 21.5 21.6 21.7 21.8 21.9 21.10 21.11

Introduction The synchrotron High quality/resolution powder diffraction Single crystal micro-diffraction Energy-dispersive powder diffraction Extended X-ray absorption fine structure X-ray microscopy Tomographic energy-dispersive diffraction imaging Conclusions Acknowledgements References

22 Electron microscopy of cements I.G.Richardson 22.1 22.2 22.3 22.4 22.5 22.6 22.7 22.8 22.9 22.10 22.11 Index

General introduction Historical context Electron microscopy Transmission electron microscopy of cement High resolution transmission electron microscopy of cement hydrate phases Scanning electron microscopy of cement Low-temperature and environmental electron microscopy of cement Analytical electron microscopy using X-rays Analytical transmission electron microscopy (TEM-EDX/EELS) Electron microscopy of cement References

457 457 458 462 467 469 469 472 477

477 477 479 482 483 494 495 495 498 498 498 500 500 500 511 511 520 523 532 533 534 540 542 557

xi

Contributors

Dr L.P.Aldridge Australian Nuclear Science and Technology Organisation Lucas Heights Research Laboratories PMB7 1, Menai Bangor New South Wales 2234 Australia

Prof John Bensted Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK

M.Attfield Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK

Prof S.Chandra Division of Applied Concrete Chemistry Chalmers University of Technology Sven Hulting Gata 8 S-412 96 Gothenburg Sweden

Stephen P.Bailey Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK Prof Paul Barnes Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK xii

Dr A.K.Chatterjee The Associated Cement Companies Ltd Research & Consultancy Directorate CRS Complex L.B.S. Marg Thane 400 604 India Dr T.M.Chrisp Department of Civil and Offshore Engineering Heriot-Watt University Riccarton Edinburgh, EH14 4AS UK Dr Sally L.Colston Industrial Materials Group Department of Crystallography Birkbeck College

Contributors

University of London Malet Street London, WC1E 7HX UK Dr D.Damidot Lafarge Central Laboratory 95 rue du Montmurier BO 15, 38291 Saint Quentin Fallavier France Dr Wolfgang Ehrfeld IMM Institute of Microtechnology GmbH Carl-Zeiss-Strasse 18–20 D-55129 Mainz-Hechsteim Germany Dr C.Famy Lafarge Laboratoire Central de Recherche 95 Rue du Montmurier—B.P. 15 38291 St Quentin Fallavier Cedex France

Dr I.Hinczak Cementech Pty. Ltd. PO Box 362 Liverpool New South Wales 2170 Australia Dr D.Hobbs British Cement Association Century House Telford Avenue Crowthorne Berkshire, RG11 6YS UK Dr S.D.M.Jacques Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK

Dr Herbert Freimuth IMM Institute of Microtechnology GmbH Carl-Zeiss-Strasse 18–20 D-55129 Mainz-Hechsteim Germany

Dr Hans J.Jakobsen Instrument Centre for Solid-State NMR Spectroscopy Department of Chemistry University of Aarhus DK-8000 Aarhus C Denmark

Dr E.M.Gartner Lafarge Laboratoire Central de Recherche 95 Rue du Montmurier—B.P. 15 38291 St Quentin Fallavier Cedex France

Dr I.Jawed Transportation Research Board National Research Council Washington D.C. 20418 USA

Prof Christopher Hall Division of Engineering and Centre for Materials Science & Engineering The University of Edinburgh King’s Buildings Edinburgh EH9 3JL UK

Dr Tom R.Jones New Materials Group Imerys Minerals Ltd John Keay House St Austell Cornwall, PL25 4DJ UK xiii

Contributors

Dr A.C.Jupe Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK Prof H.Justnes SINTEF Civil & Environmental Engineering Group Cement & Concrete Group N-7034 Trondheim Norway Prof Wies³aw Kurdowski Institute of Building Materials Academy of Mining and Metallurgy Al. Mickiewicza Pawilion A3 Krakow Poland Dr E.Lang Abteilungsleiter Forschungsgemeinschaft Eisenhuettenschlacken EV Forschungsinstitut Bliersheimer Str 62 D-47229 Duisberg Germany P.Livesey Castle Cement Ribblesdale Ltd Clitheroe Lancashire BB7 4QF UK Dr Karen Luke Now at: Halliburton Energy Services Duncan Technology Center 2600 South 2nd Street Duncan, Oklahoma 73536 USA xiv

Dr S.Lunt Thermo VG Scientific Imberhorne Lane West Sussex, RH19 1UB UK Prof F.Massazza Laboratorio Centrale Italcementi-SCF S.p.A. Via Gabriele Camozzi 124 I-24100 Bergamo Italy Dr C.E.Matulis CSIRO Coal and Energy Technology Division Lucas Heights Research Laboratories PMB7 Bangor New South Wales 2234 Australia Prof W.J.McCarter Department of Civil and Offshore Engineering Heriot-Watt University Riccarton Edinburgh, EH14 4AS UK S.Morgan Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK R.Pisula Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK

Contributors

Dr David O’Connor Industrial Materials Group Department of Crystallography Birkbeck College University of London Malet Street London, WC1E 7HX UK Prof Herbert Pöllman Institut für Geologische Wissenschaften und Geiseltalmuseum Martin Luther Universität D-06099 Halle (Saale) Germany Dr I.Richardson Civil Engineering Materials Unit School of Civil Engineering University of Leeds Leeds, LS2 9JT UK Prof K.L.Scrivener Lafarge Laboratoire Central de Recherche 95 Rue du Montmurier—B.P. 15 38291 St Quentin Fallavier Cedex France Dr Jørgen Skibsted Instrument Centre for Solid-State NMR Spectroscopy Department of Chemistry University of Aarhus

DK-8000 Aarhus C Denmark Dr G.Starrs Department of Civil and Offshore Engineering Heriot-Watt University Riccarton Edinburgh, EH14 4AS UK Prof H.F.W.Taylor Maundry Bank Lake Road Coniston, Cumbria LA21 8EW UK Dr J.C.Taylor CSIRO Coal and Energy Technology Division Lucas Heights Research Laboratories PMB7 Bangor New South Wales 2234 Australia Prof J.F.Young Center for Cement Composite Materials 205 Ceramics Building 105 S. Goodwin Avenue University of Illinois at Champaign-Urbana Urbana, Illinois 61801 USA Now at: R.D.2 Katikati New Zealand

xv

Preface

Structure and Performance of Cements was first published in 1983. Much has changed over the last 18 years, but still, a surprisingly large number of the reasons given then for the need for such a book are equally valid today; these include: • Cements constitute the second largest manufactured commodity (by weight) in the world • Most peoples’ lives are continually dependent upon the properties of cements • Only a small fraction of income derived from the manufacture and marketing of cements is spent on research into the properties and applications of cements, including new developments However, although the basic starting materials for the manufacture of most cements have not changed very much, the subject has in fact moved on appreciably. Cements today have a greater number and range of applications than ever before. The various types of cements now available (with and without inclusions of additives/admixtures) have increased significantly in numbers. New techniques of examination have arrived and our understanding of the performance requirements and utilization of many different types of cements has improved. In particular, wet process manufacture for Portland cements is declining rapidly because of the high energy costs involved in driving off the water from the raw material slurries. Modern developments now permit more environmentally friendly dry process manufacture from soft raw materials like chalk and clay with their significant moisture contents using flash calciners. Also important environmentally friendly uses for materials, previously regarded as industrial wastes in the xvi

manufacture of cements, save substantial amounts of traditional fuels like coal, gas and oil. Such pyrotechnic processes now allow wastes that used to be dumped, to be disposed off safely by burning in cement kilns on a routine basis. This is complemented by cement extension, which means that industrial by-products such as pulverized fuel ash (pfa), ground granulated blastfurnace slag (ggbs) and condensed silica fume (csf or ‘microsilica’) can partially replace Portland cement clinker in cements. Such replacement permits substantial benefits in the range of applications of the finished cements in grouts, mortars and concretes. The second edition of the book has been produced, like the first edition, for those scientists and engineers working in both the cement and general construction industry and for research and development specialists in universities and colleges, who need an up-to-date knowledge in key areas of cement technology that have moved on since 1983. It is not intended to be a standard textbook dealing with the whole range of cement types and their applications. Instead, a focused tome has been produced with contributions by a multinational consortium of authors to indicate the global developments that have arisen over the last seventeen years. Modern cement manufacturing methods, key types of cement extenders and important examining techniques, both new, and developments in existing important methods, have been highlighted for addressing a global audience. This updated book is intended to reflect both current production and research in the cement arena, but, as indicated above, is not intended to be an exhaustive treatise, which would have been unmanageably large.

Preface

The order of presentation reflects the evolutionary developments that have arisen, and is divided roughly into three broad categories: • Basic materials and methods—cement manufacture, cement phase composition, Portland cement hydration, calcium aluminate cements, properties of concrete with admixtures, special cements, various reaction/corrosion mechanisms. • Cement extenders—ggbs, natural pozzolans, pfa, metakaolin and csf—where there have been considerable developments in terms of quality and application; many other extenders have not been included where there is still much basic work needed. • Techniques of examination highlighted include the well established X-ray diffraction and electron microscopy, where there have been numerous developments in the last seventeen years, together

with more recently introduced methods (electrical impedance, NMR, synchrotron radiation-based techniques) which, with cement-based composite microstructures, need to be drawn to the attention of a wider audience. Some of the more traditional techniques have been excluded since relatively few novel applications have appeared in recent years. As the editors, we have endeavoured to obtain a widespread balance of authors from a number of different countries around the world. Many of these authors are well known internationally. They are joined by others who are newer to the field and who help to establish a broader viewpoint, both technically and through wide international coverage. We feel privileged to have succeeded in bringing such a distinguished worldwide group of authors together in one book under the umbrella title of Structure and Performance of Cements.

xvii

Notation

Standard cement chemistry notation is assumed throughout this book:

AFm denotes a solid solution range within the monosulphate-type structure (i.e. calcium monosulpho-aluminate hydrate to calcium monosulphoferrite hydrate). AFt denotes a solid solution range within the ettringite-type structure (i.e. calcium trisulpho-aluminate hydrate to calcium trisulphoferrite hydrate).

xviii

Chapter one

Cement manufacture Wieslaw Kurdowski

1.1 Introduction Cement technology is very traditional and the basic principles have remained unchanged for a long time. However, very profound changes in the production techniques were introduced leading to the diminishing of energy consumption and decrease in employment as well as fundamental pollution limitations. Especially quick development was achieved in the last thirty years, which embraced burning and grinding techniques. First, the introduction of precalciners gave a considerable increase in kiln capacity, and the specific capacity increased from 1.5 t/m3×24 h to 3.5, and even to 5 t/m3×24 h for extra short kilns. Second, in the field, the application of the roller press became a general practice and the development of roller mills is noted, which are more and more frequently used for cement grinding. The grinding efficiency was further improved by the

introduction of a new generation of dynamic classifiers. The introduction of automatic control in all production processes, including kilns, resulted in a drastic fall of employment and during weekends plants are run by two or three people. Significant progress has been achieved in pollution control. The dust emission decreased to about 0.2– 0.3 g/kg of cement. NOx, SO2 and CO are of the order 1200, 400 and 500 ppm, respectively, per cubic metre of kiln gases. For the special precalciner NOx can be as low as 400 ppm. The problem is CO2 emission, which originates mainly from CaCO3 decomposition during clinker burning. A partial solution to this problem is the increasing production of cement with additives, chiefly slag, fly ash and limestone. Also, alternative fuels are used to partially replace completely classic fuels giving a decrease of total CO2 emission from the combustion process. In the case of alternative fuels utilization, emission of dioxins and furans can be monitored continuously. The industrial experience has shown no increase in these emissions. Long-term measurements have proved that, at least in the case of some alternative fuels, the emission of hazardous gases is lower in comparison with classical fuels. Old tyres must be recognized as being number-one alternative fuels. Cement kilns are also used for the destruction of hazardous wastes instead of incinerators and the neutralization and destruction of these substances is much more effective. All these achievements make the cement industry friendly to the environment. 1

W.Kurdowski

1.2 Raw material preparation

TABLE 1.1 LSF, S R and A R of clinker produced in Germany

The properties of Portland cement is determined by the mineralogical composition of clinker. The classic Portland clinker has the following mineralogical composition: alite belite aluminate brownmillerite

55–65% 15–25% 8–14% 8–12 %

For special cements the content of C 3A and brownmillerite can differ significantly. Indispensable conditions for obtaining clinker with this mineralogical composition is the appropriate chemical composition of the bath. The share of main oxides is equal to about 95 per cent and their content is as follows: CaO

60–70% 18–22% 4–6% 2–4%

The entire content of minor components such as MgO, K2O, TiO2, Mn2O3 and SO3 is normally under 5 per cent. For the raw mix preparation limestone is used and to a lesser degree marl or clay, or even shale or fly ash. As a rule, a small quantity of the ironbearing component is also added. Sometimes, a small fraction of sand must be used to regulate the silica ratio. It is not a satisfactory solution because quartz, which is the main sand component, is very hard and remains usually in the coarse fraction and is difficult to burn. The cement plant is located near the limestone deposit which very often contains, in some deposits, rock of higher clay minerals content called ‘marl’. In cases where limestone is rich in calcium carbonate such as silica and alumina bearing components, clay is used which deposits can be found, as a rule, in the proximity of the limestone deposits. The raw mix composition can be calculated on the basis of the chosen mineralogical composition of clinker. However, to simplify the control of the chemical composition of raw meal and kiln feed, several ratios were introduced. The most important 2

*German LSF=Lime standard I=C/2.8 S+1.1 A+0.7 F.

is the lime saturation factor (LSF):

where CaO, SiO2 etc. are the oxide contents in the raw mix on the ignition-free basis. The detailed foundation of LSF is given by Lea [1]. Other ratios are as follows:

For AR lower than 0.64 the clinker does not contain the C3A phase but only C4AF and C2F. The content of the latter increases with the lowering of AR. In Table 1.1 the ratios occurring in the German cement industry are shown [2]. 1.3 Raw material crushing There have been no major changes in the technology of primary crushing of raw materials in recent years. As a rule, mainly the single- and twin-rotor hammer crushers and impact crushers are used for the primary size reduction. Jaw crushers, also in combination with roll crushers and gyratory crushers, are applied, especially for hard as well as for softer materials. For the latter, a crushing department equipped with roll crushers, arranged in two stages, is frequently used. The most recent improvements in the construction of hammer and impact crushers consist in increasing the maximum feed size and raising the

Cement manufacture

maximum feed moisture content of the raw material. For the latest construction of the twinrotor hammer crusher, the permissible feed moisture content has been doubled from 10 to 20 per cent and the mass of the individual lumps of feed material has been increased to about 7 tonnes [3]. Simultaneously, the crusher itself has a lower weight and for a unit of throughput of 1100 t/h the mass is about 160 t. In the limestone quarry, very often a mobile installation is used which provides the most economical solution. Orenstein and Koppel have furnished the mobile crusher with a very high capacity, namely 4500 t/h [3]. Depending on the terrain conditions and the required mobility, these crushers can travel with the aid of a hydraulically powered walking mechanism or can be mounted on rubber tyres or rails. The development of mobile crushing installations have been associated with a drastic weight reduction of the crusher which has become lighter by 30 per cent and more [3]. 1.4. Raw material grinding For wet and dry grinding of the raw materials, ball mills are used which operate either as open-circuit mills or in closed circuit, in case of the dry process with an air classifier. By wet grinding the process is closed through the vibrating screen or arc-screen. The quick elimination of the fine fraction in closed circuit increases the grinding efficiency by about 20 per cent. The granulometric composition of the feed greatly influences the energy consumption. To diminish the coarse fraction, the preliminary grinding is applied. For this purpose, Aerofall autogenic mills are installed which can be used for dry or wet grinding. These are one of the methods of modernization of the grinding department. The energy consumption in ball mills is also heavily influenced by the grinding media filling ratio, by the design of the mill liners and by grinding media classification in the last chamber. For the solution of the grinding media classification the so-called autoclassifying liners are used. Because of the predominance of the dry method of clinker production, in the dry grinding technology

more new achievements have been introduced. In the first place, the development of roller presses for preliminary grinding has been very often used recently. For grinding installations containing the roller press, a good solution is the application of the so-called V-separator (Figure 1.1) [4,5]. This static separator is extremely well suited for operations in grinding systems equipped with roller presses. This technology has been applied for several years in Hundai Cement Company’s Young Wol plant for cement grinding and for raw material grinding. The operation data for raw material grinding are given in Table 1.2. There are efficient installations using hammer or impact crushers for pregrinding of the material. These solutions are also convenient for raw materials drying. The hot gases from the kiln passing through the crusher transport the material to the classifier, where 70–85 per cent of the crushed material returns to the ball mill, the rest having suitable granulometry joins the product. These installations cannot be used for a very hard material because the short life of elements of the crusher can negatively influence the efficiency of the ball mill.

FIGURE 1.1 V-separator-principle scheme [5]. 3

W.Kurdowski

In the last decade, the application of roller mills increased because of the lower energy consumption of about 10kWh/t in comparison with ball mills. Large quantities of gas conveying the material to the classifier enable drying of the feed up to 20 per cent of humidity. In the past, there were some problems with the service life of the grinding elements, especially in the case of raw materials containing quartz [6]. In recent years, because of improvements in wearresistant engineering materials, the service life has been substantially increased and, depending on the abrasiveness of the raw material, values ranging from 6000 to 16,000 h are reported [3]. A special high-drying capacity characterizes Loesche roller mills [7]. These mills can work with a very high gas temperature in the range of 600°C. Practical experience gathered during the exploitation of raw materials and coal grinding can be summarized as follows: • for dry grinding of coal slurry with more than 30 per cent water content, operation temperatures of 700°C were used; • for dry grinding of raw materials with 23 per cent water content, the hot gas temperature of 560°C has been used. Among the various advantages of roller mills, the smaller dimensions and the lower capital cost (by about 10 per cent) in comparison with ball mills are worth mentioning.

The main goal of the grinding process is to ensure the suitable raw-meal granulometric composition, which usually corresponds to the residue of 1 per cent on 200 ìm sieve and 12 per cent on 90 ìm sieve. The large influence on raw-meal burnability is exerted by the coarse grains of quartz. Fundal [8] found that each per cent of silica grains larger than 44 ìm increases the free lime content of the clinker by about 0.9 per cent. So the content of quartz greater than 44 ìm should be lower than 2 per cent. Also the coarse grains of calcite, larger than 150 ìm, should be under 2 per cent. Additionally, the raw-meal moisture must be very low, as a rule below 1 per cent, commonly 0.5–0.8 per cent. The higher moisture content negatively influences the flowing properties of raw meal. 1.5 Classifiers A classifier exerts significant influence on mill efficiency and consequently on energy consumption. It is well documented by the curves showed in Figure 1.23 [47]. For the product containing 90 per cent of the fraction under 30 ìm, the energy consumption of the ball mill is 65kWh/t when the classifier efficiency is 25 per cent, but falls to 35kWh/t for 100 per cent classifier efficiency. Traditionally, two kinds of separators were used: the so-called grit separators and mechanical separators [9]. Air-swept mills are normally equipped with a grit separator and cyclone (Figure 1.2).

TABLE 1.2 Characteristic operation data: raw mill no. 3, Young Wol plant [5]

FIGURE 1.2 Scheme of the tandem drying-grinding plant. 4

Cement manufacture

FIGURE 1.3 Scheme of mechanical separator.

Grit separators have no moving parts and the separation effect is due to diminishing of dustentrained air and its tangential flow induced by guide vanes. The main part of the mechanical separator is a rotating dispersion plate and two fans: the main fan and auxiliary fan. Coarse particles either fall directly from the dispersion plate or are rejected between the auxiliary fan blades and the control valve (Figure 1.3) The fines fall into the separation volume between two vessels and drop down under inertia forces. The air stream changes the direction forcibly and returns through the vanes and auxiliary fan, forming a closed circuit. The mechanical separators have low efficiency in fine recovery, which is clearly seen on Tromp curve (Figure 1.4) [10]. To improve this efficiency a new generation of classifiers was introduced in the 1970s. The main part of this classifier is the rotor and tangential vanes (Figure 1.5). They ensure the even air velocity distribution along the full height of the rotor, which is a prerequisite for sharp separation classification. The new classifiers have a very good performance, and in comparison with classic separators produce an almost perfect Tromp curve (Figure 1.4). These separators significantly increase the mill capacity and decrease energy consumption. For example, Burgan [11] stated that the installation of a new classifier O-Sepa increased the cement mill

FIGURE 1.4 Tromp curves for classic and O-Sepa classifier [10].

FIGURE 1.5 Principle of new classifier: (1) vertical guide vanes, (2) rotor blade.

capacity by 30–32 per cent and a reduction of energy consumption by 20–30 per cent. 1.6 Homogenization Raw material homogenization is a very important technological operation because the kiln feed should have a very stable chemical composition. The variations of chemical composition of raw mix influence unfavourably the kiln exploitation and clinker quality. For a stable kiln operation, the LSF factor fluctuations should be lower than one point. Each technological operation should increase the stability of raw materials (Figure 1.6) [12]. When raw materials of a very changeable composition are used, the prehomogenization is a very effective method. However, this technology is expensive with the capital cost being about 10 per cent of total expenditure for a new plant. 5

W.Kurdowski

FIGURE 1.6 Homogeneity of raw material [12].

The effect of homogenization can be calculated using the following equation:

FIGURE 1.7 Chevcon method of stockpiling.

where Sα is the standard deviation of input variations and Sβ is the standard deviation of output variations. For estimating the variations in output, the following equation is frequently used:

There are two methods of building blending beds: longitudinal or circular. The second is better because it overcomes the so-called ‘end-cone problems’. It is difficult to keep the rate of reclaiming constant at the ends of the longitudinal pile and, especially when the reclaimer starts on the stockpile, the homogenizing effect is at first liable to be very unsatisfactory, because there will be a considerable segregation at the time of stacking. To avoid these problems, especially in longitudinal stacking, the so-called chevcon method is applied (Figure 1.7); the throw-off point of the stacker is varied by a radial distance ∆L in the course of each to and from cycle [13].

where Sx is the standard deviation of the averages of slices and N is the number of layers. In practice, there is no reason to use less than 50 or more than 500 layers (Figure 1.8) Experience shows that with the current bed blending technology, the blending effects of up to 10 are attainable by qualitatively controlled buildup of the blending bed. For monitoring of the operation of blending beds, prompt gamma neutron activation analysis can be used which was elaborated and introduced in the cement industry in the 1980s by Gamma Metrics [14,15]. This method is designated for continuous analysis of the material

6

Cement manufacture TABLE 1.3 Comparative data of various homogenizing systems [3]

FIGURE 1.8 Blending effect as a function of the number of stacked layers of material.

There are two major types of blending silo designs: turbulence and controlled flow. In a turbulence silo, the material is tumbled about by the injection of high volume air through air-pads on the silo floor. In the controlled flow system, the silo is divided into a large number of flow streams which run parallel at different flow rates, and the meal is finally blended in a tank (Figure 1.9). The fluidizing air aerates only a small part of the silo bottom. It results in low air and power consumption. The homogenizing factor for the CF FL Smidth silo is larger than seven. Modern blending silos are generally of the continuous, controlled flow type. The data for different types of blending silos after Malzig and Thier [3] are presented in Table 1.3. 1.7 Clinker burning

FIGURE 1.9 Two types of continuous homogenizing silos.

stream of grain size up to 100 mm, for example on the belt conveyor. There are several methods of raw meal homogenization [3]. Significant progress in pneumatic homogenization of raw meal has been made with continuous blending systems.

The dry method became predominant in the cement industry because of the much lower heat consumption, a typical value being 3000 kJ/kg of clinker for dry and 5500 kJ/kg for wet method. This heat expenditure for the wet method can be lowered when the water content in a slurry is under 30 per cent and the well chosen chain zone with special lifters ensures a low gas outlet temperature, not higher than 200°C. In such conditions, heat consumption can be as low as 4800–5000 kJ/kg of clinker [16]. The fuel cost can be lowered further by the use of alternative fuels. More details about this problem will be given in the point dealing with the pollution control. When the moisture of limestone in the deposit is high, for example about 24 per cent for chalk, 7

W.Kurdowski

FIGURE 1.10 Flow sheet of kiln and drying departments. Only one of two identical preheater strings is shown [20].

the wet method become competitive with the dry process. When the high humidity raw material is used, frequently the combined method of clinker production is applied. This technology consists in preparing the raw slurry in the same way as for the wet method, but then the slurry is injected into the drier-crusher and burned in the short kiln with a cyclone preheater [17] (Figure 1.10). The gas temperature is selected for the water content of the slurry through the adoption of appropriate cyclone stages (Figure 1.11). This technology was developed first in the Aalborg plant [17] and introduced successively in the Lagersdorf [18] and Rugby [19] plants. However, in the last two plants the slurry filtration in filter presses is applied and only the filter cake containing around 19 per cent of water is introduced to the drier-crusher. A very original technology was developed by F.L.Smidth in the Chelm plant in Poland [18] the chalk and marl used in this plant are very homogenous and of stable composition, being 8

FIGURE 1.11 Drying capacity of compact kiln system with calciner [20].

simultaneously very soft [21]. The raw materials grinding process is totally eliminated and driercrusher are fed directly by crushed raw materials (Figure 1.12). In this technology the splendid mixing behaviour of the cyclone preheater is utilized. This behaviour was firstly proved in the Rorhbach plant in Dotterhausen where limestone

Cement manufacture

FIGURE 1.12 Semi-dry technology without raw material grinding, Chelm case.

from the preheater was mixed with shale ash in the last preheater stage. Thermal balances of different combined technology are compared in Table 1.4 [20,21,22] The data in Table 1.4 comprise the heat consumption for the technology of filtration cake burning in the kiln with the Lepol grate preheater. This technology, that is, slurry filtration and cake nodulizing and burning in the Lepol kiln was very popular in France in the 1960s and 1970s, but totally disappeared in the 1980s. However, the dry method with granulation of raw mix and its burning in Lepol kilns does exist in France and the heat consumption is no higher than in the cyclone preheater kiln [23]. The dominating dry technology uses short kilns with cyclone preheaters. The kiln gases temperature is usually about 1100°C and the raw mix temperature about 800°C, with about 35 per cent calcination. The exit gas temperature is about 350°C for a four-stages preheater, but it can be lowered to 250°C for a six-stage preheater. The number of preheater stages is now only the question

TABLE 1.4 Thermal balances of combined methods [20–22]

a

1: Filtration, drier-crusher for cake treatment, kiln with twostage cyclone preheater; 2: drier-crusher feeding with slurry/ cake (Aalborg); 3: cake granulated and burnt in kiln grate preheater. b Water content in kiln feed 18 per cent. c Water content in kiln feed 16 per cent. d Water content in kiln feed 21.7 per cent. e Total heat consumption [20].

9

W.Kurdowski

of moisture content of the raw materials, because the gases from the kiln are used for raw meal drying. To avoid the necessity of an additional furnace for the raw mill, as a rule the four stages preheater is adopted, which ensures raw material drying of about 6 per cent humidity. In the past, the four-stage preheater was the most economical because of a high pressure drop in the cyclones. Now, new constructions of cyclones (Figure 1.13) give a lower pressure drop for six stages than before for four [24], Simultaneously, the separation efficiency of these cyclones is also higher. The examples of typical heat balances of dry kilns are given in Table 1.5.

FIGURE 1.13 Comparison of new FLS cyclone with the old form.

A great development of dry kilns was the introduction of precalcination in the 1970s. This technology results in an important increase of kiln throughput by a factor of about 2.5. Two types of precalcining technology are used; with partial and with total calcination. In the first one the air for fuel combustion in the precalciner is introduced through the kiln, increasing the coefficient of air excess, and in the other a special pipe for tertiary air must be installed (Figure 1.14). To facilitate the correct choice of precalcining technology one can consult some relations found by Herchenbach [26] and presented in Figure 1.15. The degree of calcination in the precalciner is not higher than 90–95 per cent. The unfinished calcining process constitutes a very convenient regulator preventing an excessive increase of the temperature in the precalciner. This protects against a too high temperature for the exit gases from the precalciner, and thereby, against the increase of the heat loss with the exit gases. At the same time, it protects from the possibility of a clinkering process starting in the calciner and leading to the formation of build-ups in the precalciner, which can block the inlet of the gases supplied from the kiln. The combustion of the fuel in the precalciner is carried out flamelessly and the heat exchange, which is very intensive, takes place with 90 per cent by convection. These favourable conditions of heat exchange are the reason for a very short time of

TABLE 1.5 Heat balances of dry kilns with oil or coal fired calciners in kJ/kg of clinker [16,20,25]

a Oil firing; b 6 stage preheater, calciner, modern grate cooler; c in exit gas losses; d 5 stage preheater; e total losses of cooler and clinker.

10

Cement manufacture

FIGURE 1.14 Conversion of SP kiln to an SLC system by FLS at Adelaide Brighton, Australia, the increase of production from 1800to 4100tpd [20].

calcination which is controlled by the rate of the chemical reactions. The latest achievement in burning technology is constituted by very short kilns with L/D ratio of about 10 [27–29]. These kilns ensure a very rapid clinker formation, which gives a very quick reactions run without recrystallization phenomena and, as a result, a higher hydraulic activity of the clinker is obtained. These kilns are characterized by two supports, selfadjusting roller suspensions, a tangential tyre suspension system (Figure 1.16) and a kiln drive via supporting rollers. FL Smidth also introduced a new sealing system of kiln inlet and outlet ends, a socalled double-lamella seal, which allows a higher degree of vibration at the kiln ends. The volumetric load of these kilns is of the order of 5 t/m3×24 h. FL Smidth mention further important advantages of these kilns, namely, the kiln alignment is not affected by foundation settlement [29]; also, the kiln shell ovality remains very low and the risk of kiln shell constriction is eliminated.

High volumetric loading of these kilns causes the drop in heat consumption due to a reduction of the heat losses from radiation from the kiln shell and preheater, but the main problem remains the heat losses with the exit gases. The construction of precalciners also underwent a substantial development, especially in the view of limiting pollution problems. First of all, combustion of the fuel must be very good with very low CO level in the exit gases. For this aim, the retention time of fuel in the precalciner is increased through the elongation of this furnace. It found its expression in a large volume of calciners which operate with a volume rating of 7±2t of clinker/ m3×day which corresponds to the gas retention time of 2–3 s and a meal retention time of 6–12 s [30]. The lowering of NO x emission is another important feature. For this aim, two zones in the calciner are created: a reducing zone in the lower part, and an oxidation zone in the upper part of the precalciner [30–32]. In the first zone, the following reactions occur:

These reactions are favoured by a high temperature and catalysed by the raw meal. In the upper part, the remaining CO is oxidized to CO2, but because of the low temperature in the precalciner no additional oxidation of nitrogen will take place. A very interesting example of a low CO and NOx calciner is the ILC low-NOx calciner of FL Smidth [31]. In the inlet of the calciner, the reduction zone is created where coal is introduced and the only oxygen available is the amount present in the rotary kiln gases. This favours NOx reduction, according to the previously mentioned reactions. The high temperature that improves the efficiency of the first reaction is provided by splitting the raw meal between the oxidizing zone and the reduction zone. The temperature is controlled to 925–1050°C, or as high as possible without getting any encrustations in the kiln riser 11

W.Kurdowski

FIGURE 1.15 Relation between clinker production, kiln diameter, heat input in preheater, burning zone rating, gas velocity in kiln inlet and tertiary air control [26].

FIGURE 1.16 Plan view of FLS tangential tyre suspension system [29].

duct and reduction zone. Typically, 15–30 per cent of the raw meal should be fed to the reduction zone inlet (Figure 1.17). It is very important to assure the strongest possible reducing conditions in order to 12

favour the reactions of NOx decomposition. In the ILC low-NOx calciner, all the fuel is introduced to this zone before the tertiary air. The result of joint effects of all the decomposition reactions is the removal of approximately 70 per cent of the rotary kiln NOx. The restriction in the middle of the calciner and the bend in the top are applied in order to improve the mixing, thus minimizing the CO emission and at the same time improving the fuel burnout. Even though after the reduction zone more than 3 per cent CO is in the gas, the concentration of the CO in the calciner exit gas will only be around 100–200 ppm, due to these facilities. A very similar principle was adopted in the Pyrotop calciner of KHD Humboldt [30] (Figure 1.18). The kiln exit gases and tertiary air are

Cement manufacture

FIGURE 1.18 Pyrotop calciner of KHD Humboldt [30]. FIGURE 1.17 ILC low-NOx calciner of FLS [31].

introduced into the calciner at an acute angle, which results in the slow mixing of the two flows. The fuel is burnt at two different locations and also the meal is admitted at two different points. The calciner is equipped with a new reaction chamber mounted at the location of the 180° elbow. It ensures a reliable mixing of the flow portions originating from the kiln waste-gas duct and from the tertiary air duct. Furthermore, a portion of the meal is discharged from the chamber and may either be recycled, that is, be routed back to the ascending calciner section, or directed to the calciner section running downward. A repeated deflection of gas/meal flow especially achieves an intensive mixing and thereby enhances combustion. The turbulent flow at the calciner outlet is a decisive prerequisite for good burning out of the fuel [30]. The construction of burners also underwent substantial changes and instead of very simple

burners, the multi-channel burners were introduced [32–34]. Pillard was among the first people to develop this type of burner, which was adopted for a simultaneous use of several fuels [34]. The new constructions eliminate a very low fuel/air ratio, which was typical for old coal burners, and result in the higher combustion efficiency and a greater possibility for flame regulation. The volume and the shape of the flame may be changed by adjusting the ratio between the radial and axial air amount and velocity by adjusting the burner nozzle areas (Figure 1.19). Special constructions which diminish a NOx formation in the flame, have been very popular recently [35]. 1.8 Coolers Clinker cooling is very important for heat consumption in the kiln and for cement quality. This process is conducted in the coolers. There are three types of coolers which are generally used: rotary coolers, planetary coolers and grate coolers (Table 1.6). 13

W.Kurdowski

FIGURE 1.19 Multi-channel burner. TABLE 1.6 Data of clinker coolers for dry kilns [40–43]

a

FLS Unax-cooler. Cross-sectional ratings, inclination 4.5–5 per cent. c Specific area rating. d In hot climatic conditions, about 20kJ/kg of clinker. e Above ambient temperature. f With tertiary air. b

14

Cement manufacture

Grate coolers are most common because they ensure the lowest clinker temperature and highest kiln capacity [16,36]. Planetary coolers take second place—they are mounted on the kiln, and the air for cooling is provided by the kiln fan. Rotary coolers are seldom used. The superiority of grate coolers is finally caused by the convenience of tertiary air supply to the calciner. The effiency of rotary and planetary coolers is closely connected with the heat transfer, which is enhanced through a geometric configuration of the cooler and arrangement of internal fittings. A special system of elbows and different lifters, both for planetary and rotary coolers, has been developed by Magotteaux and Estanda [37,38]. Planetary and rotary coolers have a limited airflow, usually 0.85–0.95 nm 3/kg of clinker, depending on the fuel consumption of the kiln and regulated by the kiln ID fan. For these reasons, the clinker temperature exceeds 150°C and water injection is applied. On the contrary, the quantity of air in grate coolers, which is supplied by special fans, is much higher, up to 2.5 nm3/kg of clinker and the clinker temperature is low, usually 50°C+ambient temperature. However, the wellknown problems with exit air must be resolved. The best solution is to use the exit air for drying raw materials, slag and coal. In some cases electrical energy is produced [39]. For dedusting of exit air, the electrofilter or gravel bed filter is used. To diminish the quantity of cooling air different, variants of air recirculation have been employed and the exit air of low temperature, that is, 250°C, is reintroduced to the hot part of the cooler [39]. The grate cooler is frequently equipped with the roll crusher, which can work at a high temperature [41]. It is very important because the clinker cooling rate depends strongly on the grain size of this material. The grate coolers regularly underwent developments in construction. The construction of the plates, and air supply directly to the plates. in the so-called air-beam technology (Figure 1.20) have been recent achievements. The new plates have narrow air outlet slots, which ensure a sharp air

FIGURE 1.20 Aeration system for directly aerated troughed plates [41].

jet of about 40 m/s in the direction of clinker transport. This solution permits the specific loading of the cooler to be raised, even the level to 100 t/m2×d has been tested [41]. The Coolax of F.L.Smidth [40] is an example of a new cooler. In the heat recuperation zone, cooling air is supplied directly to the grate plates through a system of ducts and hollow beams instead of via an undergrate compartment as in conventional coolers. This system also eliminates the so-called ‘red river’ phenomena in the cooler. The principal Coolax features are: • fuel savings of 130–170 kJ/kg of clinker; • 30 per cent less cooling air and 40 per cent less air to be dedusted; • smaller overall dimensions due to the high specific grate load-up to 60 t/m2×24 h. A very original construction was developed by Smidth and Fuller [42,44]. It is a so-called SF CrossBar Cooler in which the cooling process is separated from the conveying action. A large drag chain pulls the clinker over a stationary bed through which cooling air is blown. Each grate plate is equipped with a rectangular tower containing a weighed mechanical flow regulator, which acts as a barometric damper in a chimney (Figure 1.21). All these solutions enable a decrease of the cooling air input to only 1.5–1.7 nm3/kg of 15

W.Kurdowski

clinker. The heat losses are also diminished to 400 kJ/kg of clinker. Environmental protection regulations limiting dust emission have led to developments in construction of the second degree coolers working in conjunction with a short grate cooler [45]. The latter supplies hot air to the kiln, and the clinker at approximately 500°C, after being crushed in a breaker, is fed to a static membrane g-cooler (Figure 1.22). In this cooler the clinker, descending slowly at the speed of 2–3 cm/min, is cooled down to

FIGURE 1.21 Aerating system for plates in SF Cross-Bar Cooler [42]. (1) air flow through holes in tongue; (2) air flow through idle hole; (3) air flow through gaps.

FIGURE 1.22 Short grate cooler used in combination with a gravitation cooler [45]. 16

100°C, heating the air flowing in the tubes up to 60–100°C. The specific capacity of this cooler amounts to 10 t/m3×24 h. The specific power demand of the static cooler is about 1.2kWh/t of clinker and of the whole installation is about 5kWh/t. The gcooler has been applied at the Robilante plant [46], for example. 1.9 Cement grinding For cement grinding, ball mills are more frequently used. They have a rather high-energy consumption, typically 34kWh/t for cement having a specific surface of 300 m2/kg. For good operational results, the level of ventilation, typically 0.3–0.5 nm3/kg clinker and volumetric charge loading which is 25–30 per cent for the first and 26–28 per cent for the second compartment, are very important. The modern mills are two chamber mills with lifting liners in the first and classifying liners in the second compartment. The diaphragms play also a very important role. The new construction of Slegten-Magotteaux or Combidan of F.L.Smidth provide a good ratio of the mass of material to the mass of the balls and thus improve the effectiveness of grinding [43]. Much better results are achieved by the mills working in closed circuits which have lower energy consumption and higher capacity of 20 per cent in comparison with the same mill working in an open circuit. The efficiency of classifiers is also important. This is presented in Figure 1.23 [47].

FIGURE 1.23 Influence of classifier performance on mill power consumption [47].

Cement manufacture

Ball mills have low efficiency, the majority of energy being transformed into heat. Significant progress in grinding technology has been due to the application of roller press for pregrinding or finish grinding [48]. The material between the rolls is submitted to a very high pressure ranging from 100 to 200 MPa. During the grinding process, partial breakdown of the grains occurs and there is a partial formation of incipient cracks in the interior of the grains. The material leaves the press in the form of agglomerates. When the roller press works in the closed circuit for finish grinding, a disintegrator must be used or a special separator must be installed to deglomerate the product. Frequently the roller press is applied for pregrinding with the ball mill as a second stage, thus the rise in capacity is of 30–60 per cent and 15–25 per cent for economy of energy [49]. Patzelt [49] gives, as an example, the installation of the roller press (dimensions 1.7× 0.63 m) with a motor 2×575kW which raised the raw mill throughput from 190 to 280 t/h that is, by about 47 per cent. Regardless of the fact that special hard materials are used for the lining of the rollers, the short-coming of this technology is relatively low reliability. It is due to the fact that the rollers are operating at a very high pressure. The maximum local pressures are near 350 MPa for clinker grinding [50]. It was the main reason for further development in the construction of special mills, and horizontal roller mills were produced. The so-called Horo-mill (Figure 1.24) was the first with a grinding pressure evidently lower, that is, 40–70 MPa. These mills are working in several cement plants in Europe, Turkey and Mexico [51,52], having a capacity up to 100tph. The energy consumption of the mill alone is about 16kWh/t. Another horizontal mill, the so-called Cemax, was developed by F.L.Smidth [53]. The Cemax grinds the material by compressing it between a ring and a roller. The energy consumption of the installation is about 25kWh/t and is similar, both in the Horo-mill and in the Cemax. The quality of cement from the Horo-mill, Cemax and roller press is also quite similar and

FIGURE 1.24 Principle of the Horo-mill.

comparable to the product of ball mills. A little higher water demand for the product of the roller press is noted. Experience with roller mills used for cement grinding grows constantly. Ten years’ experience in exploitation of a Pfeiffer roller mill in the Teutonia Zementwerk AG near Hanover has proved that these mills can be used for cement grinding [54]. The slag cement can be ground without preliminary slag drying. A specific power consumption by the grinding plant in the range 21– 33kWh/t can be attained. Thus the economy in energy, in comparison with ball mills, is between 25 and 45 per cent. At the end of the 1970s, Loesche [55,56] started to develop new technologies in vertical roller mills, the so called 2+2 and 3+3 technology, which were applied with great success for cement and slag grinding. The main difference between the 2+2 mills and the conventional four-roller-mills for raw material grinding is that two differently sized pairs of rollers are used (Figure 1.25). The so-called srollers are used to prepare the material bed on the grinding table. These s-rollers are of light design and have a smaller diameter compared to the large grinding rollers, the so-called m-rollers. By means of 17

W.Kurdowski

FIGURE 1.25 Diagramatic representation of grinding bed preparation [55].

hydraulic attachments the small, light s-rollers are positioned at a certain height above the grinding table, where they slightly consolidate and deaerate the material bed. After this preparation the material is properly nipped under the large and heavy grinding rollers of conventional design. The forces required for comminution underneath the m-rollers are created by means of hydraulic cylinders attached to the rocker arms of the m-rollers. All m-rollers are supported on individual stands, which are relocated on the mill foundation. Furthermore, individual hydropneumatic spring systems for each roller provide a very smooth and almost vibration-free operation for the mill. The 2+2 system is available for capacities of up to 170tph OPC. In case of higher throughput the 3+3 system is used. Service lives of the roller shells, without any intermediate reworking, of more than 8000 h, have been achieved in cement mills in Japan and Korea and up to 15,000 h have been achieved for grinding track plates. The specific energy consumption for a complete cement grinding plant shown on Figure 1.26 is between 26 and 29kWh/t for a finished product fineness, corresponding to Blaine’s value of 3,300cm2/g. The comparison of energy consumption for grinding slag in a ball mill and Loesche mill is given in Table 1.7. There is the Hon Chon of the Morning Star Co. cement plant in Vietnam, in which only roller mills are used and no ball mill is installed [57]. 18

FIGURE 1.26 Cement grinding plant with roller mills [55]. TABLE 1.7 Power consumption for grinding of 60tph blast furnace salg 4300cm2/g [55]

1.10 Automation The process technology and mechanical equipment for cement production has became increasingly complex in recent years. Product quality and diversity, environmental protection, secondary fuels, and energy management require more sensitive plant management. These are the main reasons for the introduction of computers for process control and for information systems. The classical structure of the process control and information systems is distinguished by a relatively clear pyramid hierarchy (Figure 1.27) [58]. There is a constant increase in information density which affects every level in the computer integrated manufacture pyramid. Hardware and software with appropriate functions are installed at different levels of the computer integrated manufacture pyramid and linked with one another through communication

Cement manufacture

FIGURE 1.27 Structure of the process control and information systems (CIM pyramid) [58].

buses to an integral process control and information system [58]. Nowadays, process control and information systems are not just a means of operating production plants; they are integrated parts of the plant and are essential for efficient, low-cost manufacture of quality products and for effective plant control and monitoring. Many well-functioning computer-based systems for supervision, dialog and reporting have been introduced by different cement equipment producers as well as cement producers. Among others the HLC of Holderbank [59], TOPEXPERT of Ciments François [60], SDR of FL Smidth [61], PRODUX of KHD and POLCID of Krupp-Polysius are the most popular ones [58]. The task of the higher-ranking process computer is to determine control inputs for the process and make them available for process control, to provide long-term storage of the process data and put them out as listings or diagrams, and also to make possible the dialogue between the operating personnel and decentralized processors with maximum convenience [62]. For many years, the automation concepts for raw meal preparation have been successfully introduced into cement manufacture. It was a very important step towards achieving a uniform kiln operation. But the problem of identifying a suitable microprocessor, based on the supervisory kiln control technique, took longer to resolve. Therefore, several methods were developed which were based on modelling of the operator actions

[62–64]. Basically, the technique involves expressing ill-defined or ‘fuzzy’ operator actions or commands in terms of precise mathematical logic. The fuzzy logic kiln control was applied to hundreds of kilns and the result fully proved its usefulness for automatic control of the burning process [65–67]. The introduction of automatic control of kilns saves thermal energy by 3 per cent, electrical energy by 2 per cent, results in production increase by 5 per cent, consumption of refractory lining diminution by 6–10 per cent and variance of clinker quality diminution of 4 per cent [60]. In case of mills the production increase is of up to 10 per cent, the decrease of energy consumption is 7 per cent, and, simultaneously, the quality variations are eliminated. The actual breakthrough in the automation of rotary kilns and also the mills appeared with the introduction of new heuristic control methods, such as fuzzy logic, rule-based logic and online expert systems, coupled with the availability of powerful computers. In Europe alone there are now more than six suppliers of expert systems for cement kilns and ball mills [53]. In the last twenty years, new and very important measurement techniques were introduced. Perhaps the adoption of NOx measurements to follow the temperature level in the sintering zone is most important [68]. Another very important tool in cement kiln control is the automatic supervision of protective coating thickness and hot spots on the kiln surface through kiln shell temperature measurements by the infrared technique. Several solutions of this problem are proposed by cement equipment producers [69]. Also the laser technique for cement fineness control or free lime control in clinker, which was introduced by Lafarge, is worthy of mention. Laboratory automation has also become extremely important recently. Its particular function is to ensure that the data from raw materials and the products relevant to quality are continuously available to ensure a quality-driven process control. Full automation of laboratory functions is supplied by different producers. As an example, the Polysius system POLAB 3 can be mentioned which uses modular technology and 19

W.Kurdowski

includes two completely separate sample preparation lines with five robots, including even the automatic fusion equipment [70]. The entire system normally comprises a representative sampling system, a plant for rapid sample transport, a sample processing system and analysis technology suited to the special features of the material samples, and reliable open- and closedloop control equipment for ensuring and monitoring the quality [71]. For the raw materials, chemical composition control, very great progress can be achieved through prompt gamma neutron activation analysis [14,15] which was mentioned in point 3. 1.11 Environmental protection The manufacture of cement necessarily involves interference with the environment, although the cement industry is doing its best to minimize the negative effects of this. The problem comprises the water discharge, solid waste, noise and air pollution. Water discharge should not be a concern beyond handling normal domestic waste and storm water run-off with its potential for leaching from stockpiles and spillage. Solid waste practically concerns kiln bricks. When chromium bricks are eliminated, the alumina bricks can be added to the raw material and magnesium bricks sent to the landfill. Air pollution has been the main problem for the cement industry and will be discussed in more detail. For twenty years, a significant progress has been made within the cement industry in the field of air pollution protection. The dust emission problem is now practically solved and the new construction of electrostatic precipitators and fabric filters, operating with a compressed air (reverse-pulse) cleaning system [72], have ensured a very low emission level, typically under 50mg/nm3 of gases. However, for good electrostatic precipitator efficiency, the dew point of the gases is very important and for the dry method is always too low. It affects negatively the dust resistivity (Figure 1.28) Thus the conditioning towers in the gas circuit are installed and are in operation when the grinding/drying unit is not working. 20

FIGURE 1.28 Resistivity of kiln dust as a function of temperature for different moisture levels of kiln gases (1.3, 5, 10 and 15 per cent).

Electrostatic precipitators underwent substantial modernization in 1990s and all modifications are reviewed in Peterson’s paper [73]. Maybe the most important is the pulse energization system. By superimposing short-duration high voltage pulses on DC voltage, the charging of particles as well as the collecting efficiency of the precipitator are improved. The improvement factor of the precipitator performance is defined as a ratio between the dust migration velocity value wk for pulse and for DC energization ascertained when both are adjusted to the optimum performance [74]. The wk concept is useful since the required precipitator volume is inversely proportional to this value (Figure 1.29). Precipitator efficiency and the wk migration velocity can be determined from the inlet and outlet dust concentration and gas volume flow rate:

Cement manufacture

FIGURE 1.29 Electrofilter efficiency as a function of A/Q for three different values of wk.

where wk=particle migration velocity, m/s; Q=actual gas volume flow rate, m3/s; A=total collecting area, m2; k=constant (0.5); η=collecting efficiency, %. The efficiency as a function of A/Q is shown for three different values of wk in Figure 1.28. In pulse energization the discharge current can be regulated independently of the precipitator voltage by variation of the pulse repetition frequency and the pulse height (Figure 1.30). This makes it possible to reduce the discharge current to the threshold limit for the back corona with high resistivity dust without reducing the precipitator voltage. The gaseous emission comprises NOx and SO2, and in the case of alternative fuels, dioxins and furans. The nitrogen oxides are formed mainly in the burning zone and are mostly thermal NOx [75]. The typical level of NOx for a cement rotary kiln is 1000–1500 ppm in which 90 per cent is thermal NOx. There is also a fairly stable 200–250 ppm background contribution of NOx associated with fuel nitrogen and this is generated at relatively low temperatures within the flame itself. In the kilns with calciners, the NO x content becomes considerably lower because of the low burning temperature in the calciner. Also, the special

FIGURE 1.30 Current density as a function of repetition frequency for different Up values.

deNOx calciners are used which were described earlier. In the case of these calciners only 0.35 g/kg of clinker comes from the kiln and around 0.7 g NO2 from the calciner, so that only around 1.05g NO2/kg of clinker is emitted, which is a little lower than 500 mg NO2/nm3 (10 per cent of oxygen 21

W.Kurdowski

level). Because there is no ascertainable elimination of NO x in conditioning towers or grinding/drying plants, when the decrease of NOx in exit gases is indispensable, a very promising method is a non-catalytic decomposition of NO with NH3 [76]. The selective reaction of NH3 with NO takes place in the presence of oxygen, preferably at temperatures of about 950°C. Kreft [77] determined the best location of NH3 injection to the cyclone preheater (Figure 1.31). The emission of sulphur usually presents no problems. Additionally, the emission of SO2 can be reduced by passing the gas through a grinding mill or conditioning tower. The cement burning process and grinding process thus function as ideal desulphurizing systems in which more than 90% of SO2 is retained [72]. However, in some cases, when sulphides are present in raw material, the problem must be specially solved, because sulphur is combusted in the preheater and SO2 leaves the kiln with the exit gases [78]. Bonn and Hasler [78] present the opinion that only absorption in a circulating fluidized bed can promise sufficient lowering of the SO2. In the fluidized bed the exhaust gas is brought into contact with water to lower the temperature down to the dew-point range, then with a mixture of raw meal and slaked lime, the SO2 is combined as sulphite and the product is fed back to the kiln, though in limited quantities. A very efficient desulphurization system is also in use at the Aalborg cement plant [79]. Polysius implemented an innovative system, called POLVITEC, which solved the problem of high SO2 and NOx emission and simultaneously provided

the preconditions for environmentally acceptable utilization of sewage slurry as a fuel [80]. The system comprises an activated coke-packed bed filter in combination with an ammonia liquid injection into a kiln inlet. The pollutants contained in the kiln exhaust gases, including sulphur dioxide, ammonium compounds, heavy metals, hydrocarbons and residual dust from the upstream electrostatic precipitator are collected in the activated cokepacked bed filter. The gas passes the filter in the cross-current mode. The activated coke moves vertically through the housing. Collecting the replacement of activated coke takes place discontinuously in small steps. The exhausted activated coke is used as a fuel in the kiln. Cement kilns are very promising reactors which effectively burn even hazardous waste. Because of the high flame temperature, the air excess cement kiln presents excellent conditions for burning of the most difficult organic compounds, that is PCB and PAH, so all hazardous wastes are destroyed and neutralized [81]. Simultaneously, a strongly basic medium in cement kilns absorbs the heavy metals and the immobilization is of the order of 99.99 per cent [82]. As long practice has shown, the emission of toxic substances and heavy metals from the cement kiln is as a rule lower than in the case of classic fuels [82]. However, the quantity of noxious components in the alternative fuels must not exceed the threshold quantity, which is shown in Table 1.8 [82]. The use of alternative fuels has been developed especially in France where the share of these fuels reached 52.4 per cent in 1994 [83]. The different

FIGURE 1.31 Reduction of NOx for two location of feeding points in the suspension preheater. 22

Cement manufacture TABLE 1.8 Pitch acceptability limits [82] in mg/kg of fuel

combustible wastes which are used as fuels are: petroleum coke, used lubricants, oils, paints, organic solvents, tar, waste-containing cyanides, pesticides, sundry organic and inorganic and others. The most popular alternative fuel for a long time has been old tyres. 1.12 Acknowledgements The author wishes to express his gratitude to F.L.Smidth for sending very useful technical papers and especially to Mr. Richard Eimert and Bo Bentsen for their kind help. The author would like to thank Professor Bensted for his careful correction of the manuscript. Special thanks are due to Mr. Roland Nizecki for technical help. 1.13 References 1. Lea, F.M., The Chemistry of Cement and Concrete, 3rd edn., Chemical Publishing Company, New York (1971). 2. Sprung, S. and Rechenberg, W., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 121 (1994). 3. Malzig, G. and Their, B., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 178 (1987). 4. Strasser, S., Cement-Wapno-Beton (in Polish), 63, 171 (1996). 5. Schultz, A. and Lube, B., World Cem., June, p. 28 (1997). 6. Duda, W., Cement Data Book, Bauverlag GmbH, Wiesbaden, Berlin (1985). 7. Brundiek, H. and Poeschl, H.J., Cement-WapnoBeton (in Polish), 65, 56 (1998). 8. Fundal, E., World Cem. Technol., 10, 604 (1979). 9. Alsop, P.A. and Post, J.W., Cement Plant Operations Handbook, 1st edn, Tradeship Publications Ltd (1995). 10. Knoflicek, VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 222 (1987).

11. Burgan, J.M., Zem.-Kalk-Gips, 41, 350 (1988). 12. Jung, C., Int. Cem. Rev., June, p. 71 (1999). 13. Frommholtz, W., Aufbereitung-Techn., 22, 119 (1981). 14. Poleszak, J., Cement-Wapno-Gips (in Polish), 60, 189 (1993). 15. Tschudin, M., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 259 (1994). 16. Kurdowski, W., ‘Clinker burning technologies’, in Advances in Cement Technology, S.N.Ghosh (ed.), p. 68, Pergamon Press, London (1983). 17. Borgholm, H.E. and Nielsen, P.B., World Cem., 20, 74 (1989). 18. Hargreaves, D., Int. Cem. Rev., June, p. 28 (1999). 19. Jackson, P., Int. Cem. Rev., June, p. 30 (1996). 20. Palle, Grydgaard, ‘Kiln conversions’, FL Smidth (1998). 21. Kurdowski, W., 3rd NCB Int. Sem. Cem. Build. Mater., New Delhi, Vol. 2, p. III-29 January (1991). 22. ‘Energy saving in cement manufacturing’, FL Smidth Technical Literature, LST/RS/H 87–3A–4177. 23. Erhard, H.S. and Scheuer, A., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 278 (1994). 24. Igawa, T. and Hanato, H., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 420 (1987). 25. Kwech, L., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin p. 430 (1987). 26. Herchenbach, H., Zem.-Kalk-Gips, 34, 395 (1981). 27. Mrowald, U. and Hartmann, R., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin p. 416 (1994). 28. Weber, H., World Cem., 13, 255 (1982). 29. ROTAX-2, FL Smidth Technical Data. 30. Bauer, C., KHD Humboldt Symp, p. 67 (1990). 31. Smidth, F.L., ‘ILC low NOX calciner’, FL Smidth Technical Literature. 32. DUOFLEX burner, ‘Swirlax kiln burner’, FLS Technical Literature. 33. Steinbiss, E., Bauer, C. and Breidenstein, W., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 454 (1994). 34. Pillard Technical Literature. 35. Thomsen, K., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 624, (1994). 36. Herchenbach, H., Zem.-Kalk-Gips, 31, 42 (1978). 37. Fonderies Magotteaux Technical Literature. 38. Estanda Technical Literature. 39. Nakamura, N., 7th Int. Con. Chem. Cem., Paris, Vol. IV, p. 725 (1980). 40. Coolax grate cooler, FL Smidth Technical Literature. 41. Steffen, E., Koeberer, G. and Meyer, H., VDZ 23

W.Kurdowski

42. 43.

44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58.

59. 60. 61. 62. 63.

24

Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 449 (1994). Bensten, B. and Keefe, B.P., Zem.-Kalk-Gips, 52, November, p. 608 (1999). Kurdowski, W., ‘Cement manufacture’, in Cement and Concrete Science and Technology, S.N.Ghosh (ed.), Vol. 1, Part I, ABI Books, New Delhi (1991). Keefe, B.P. and Puschock, E.L., World Cem. p. 52, January (1999). Buzzi, S. and Sassone, G., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 296 (1994). Wedel, K., Zem.-Kalk-Gips, 38, 87 (1985). Cleeman, J. and Modweg-Hansen, C., Zem.-KalkGips, 27, 337 (1974). Wustner, H., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 279, (1987). Patzelt, N., Tonind.-Ztg., 113, 383, 662 (1989). Cordonnier, A., Zem.-Kalk-Gips, 47, 643 (1994). Buzzi, S., Zem.-Kalk-Gips, 50, 127 (1997). Int. Cem. Rev., February, p. 30 (1995). CEMAX, FL Smidth Technical Data. Leyser, W., Hill, P. and Sillem, H., Ciment, Bétons Platres Chaux, 761, 200 (1986). Brundiek, H., Cement-Wapno-Beton (in Polish), 65, 42 (1998). Schafer, H.U., Cement-Wapno-Beton (in Polish), 65, 48 (1998). Schafer, H.U., Personal Communication. Sauberli, R., Herzog, U. and Rosemann, H., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 148 (1994). Hasler, R. and Dekkiche, E.A., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 216 (1994). Kaminski, D. VDZ Kongress 1993, Bauverlag GmbH , Wiesbaden, Berlin, p. 225 (1994). Holmblad, L.P., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 539 (1987). Holmblad, L.P., Zem.-Kalk-Gips, 39, 192 (1986). Munk, R. and Ruhland, W., VDZ Kongress 1985,

64. 65.

66. 67. 68. 69.

70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82.

83.

Bauverlag GmbH, Wiesbaden, Berlin, p. 504 (1987). Haspel, D.W., Taylor, R.A., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 530 (1987). Holmblad, L.P. and Oestergaard, J.J., Control of a Cement Kiln by Fuzzy Logic, North-Holland Publishing Company, Amsterdam (1982). Oestergaard, J.J., ‘Fuzzy control of cement kilns’, FL Smidth Technical Literature. Bentrup, K., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 600 (1987). Lowes, T.M. and Evans, L.P., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 305 (1994). Pisters, H., Becke, M., Spielhagen, U. and Jager, G., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 460 (1987). Triebel, W., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 255 (1994). Teutenberg, J., VDZ Kongress 1993, Bauverlag GmbH, Wiesbaden, Berlin, p. 252 (1994). Kroboth, K., Xeller, H., VDZ Kongress 1985, Bauverlag GmbH, Wiesbaden, Berlin, p. 600 (1987). Petersen, H.H., ‘New trends in electrostatic precipitation’, FLSmidth-Review no. 35, 1980 Coromax, upse energisation. FLSmidth-Newsfront, 1983 Haspel, D.W., World Cem., 14, 176 (1989). Scheuer, A., Zem.-Kalk-Gips, 43, 1 (1990). Kreft, W., Zem.-Kalk-Gips, 43, 153 (1990). Bonn, W. and Hasler, R., Zem.-Kalk-Gips, 43, 139 (1990). World Cem., Anon., 54 (1998). Rother, W., Int. Cem. Rev., January, p. 39 (1996). Xeller, H., Cement-Wapno-Beton (in Polish), 64, 223 (1997). Salvarode, D. and Zappa, F., ‘Improved technologies for rational use of energy in the cement industry’, Proceedings, Berlin, October 26–28, p. 409 (1992). Cembureau, Eur. Annu. Rev., 17, XIII (1994).

Chapter two

Composition of cement phases Herbert Pöllmann

2.1 Introduction Cement components are usually described by their phase relations out of phase diagram determinations. Basically, the most important cement phases can be shown in the phase diagram CaO–Al 2O 3–SiO 2 (Figure 2.1) with additional MgO and Fe 2 O 3 and some minor, but also important components. Different hydraulic and latent hydraulic materials are shown in Figure 2.2. The stability fields of the phases can be given by these diagrams, but due to the multicomponent system, it is often difficult to describe solid

FIGURE 2.1 Hydraulic and latent hydraulic materials in the system (CaO+MgO)–(Al2O3+Fe2O3)–SiO2.

solutions, phase transitions, phase stabilization and true element contents. For that, often isoplethic or isothermal sections are used to show the details. For these already quite well-known phase diagrams, the phase boundaries, melting points, 25

H.Pöllmann

FIGURE 2.2 The ternary phase diagram CaO–Al2O3–SiO2 showing the three binary phase diagrams CaO–SiO2, CaO–Al2O3 and Al2O3–SiO2 (Bowen, N.L. and Greig, J.W. 1924; Greig, J.W. 1927; Phillips, B. and Muan, A. 1959; Rankin, G.A. and Wright 1915; Roy, D.M. 1958; Welch, J.H. and Gutt, W. 1959).

transition points and stability regions can be given in detail. In a “cement system” characterization of compositional changes, solid solutions and the additional knowledge of the influence of minor contents of elements as a function of temperature can lead to differences in non-equilibrium systems. The existence of non-equilibrium conditions can often be seen by zoned crystals in cement clinkers. The complex phase composition of cement minerals is also now increasingly influenced by the use of alternative fuels, waste compounds and fluxing agents which can cause phase changes or 26

formation of new phases. Besides, for phase diagrams which are based on pure components, the in situ determination of compositions of phases is possible by using micro-techniques such as microdiffractometry, EPMA or EDX analysis. The varying chemical contents and impurities, the crystallographic changes of clinker minerals coming from different sources, kilns and production techniques can be measured and shown directly. Due to these varying conditions, the equilibrium mineralogical calculations such as

Composition of cement phases

Bogue or modified Bogue are increasingly replaced by direct quantification methods to describe the real phase assemblages quantitatively. As microscopic techniques are mainly used to determine quality and quantity of clinker minerals at moderate times, the X-ray Rietveld technique can be developed for different clinkers even as a completely automized method for quality control. The quantification results can be obtained in at least 10–20 min after a proper installation of the control files. Phase compositions, crystal chemistry and crystallographic parameters of all phases must therefore be known. The next steps will therefore provide direct determination methods qualitatively and quantitatively. A comparison of the advantages and disadvantages of microscopy, X-ray methods and Bogue calculations are summarized in Table 2.1. Small amounts of additional phases which can occur due to increasingly used impure raw and secondary materials cannot be found directly by Xray methods, but must also be identified by different additional methods. Despite the knowledge of phase formation and crystallization conditions derived from phase diagrams, a description and investigation of clinkers and phase formation requires different methods: 1. Bogue calculation of theoretical phase contents including modified Bogue. 2. Direct X-ray analysis. 3. Rietveld analysis. 4. Chemical methods like free-lime determination. 5. X-ray fluorescence. 6. Electron microprobe analysis for small-sized samples and individual chemical information. 7. Optical methods like reflected-light microscopy, powder mounts and thin sections. 8. Scanning electron microscopy with fluorescence, secondary and back-scattered images. 9. Selective dissolution methods for enrichment of phases. 10. Thermoanalytical methods for alkali phases and carbonization effects. 11. Spectroscopic methods. 12. EPMA, ESR, ICP, AAS.

2.2 Simple oxides, elements and their potential role in cement chemistry The role of simple oxides in cement chemistry is dominated by the free lime content which is measured in cement plants for quality control. High free lime contents can be an indication that burning conditions or homogenization are not successfully adapted. Low-burnt clinkers can have increased content of free lime. With grinding and transport of fine ground material, often carbonization can take place. This must be taken into account when free lime determinations are made routinely after some time, because lower free lime contents are measured. Free lime and free magnesia contents of clinkers are restricted due to the hydration of phase on-expansion reactions which can occur in the final product. Other oxides like SiO2 can be due to coarse quartz in the raw meal or due to incomplete homogenization. Corrective additives like ironcontaining compounds can lead to the formation of iron spinels like magnetite or magnesium containing spinels. Spinels may occur due to reactions of metals2+ and metals3+. Other iron oxides like wüstite were identified when reducing conditions occured in the clinker. The incorporation of Fe 2+ in alite can lead to a destabilization of alite. With increasing reducing conditions, even native iron was found in clinkers. Free titanium dioxide is normally not found in clinkers, but can be found when increased amounts of painting residues are added and incomplete homogenization takes place. In Figures 2.3 and 2.4, free lime and an iron-rich spinel coming from reducing conditions are shown. Some details for these phases are summarized in Table 2.2. 2.3 Calcium silicates (CxSy) The system of CaO–SiO2 was investigated by Welch and Gutt (1959), Philips and Muan (1958) and Roy (1958). Four binary phases exist within the system which are increasingly hydraulic with increasing C/S-ratio. The different modifications of 27

28

TABLE 2.1 Comparison of Bogue, microscopy and X-ray techniques for phase determinations (Pöllmann et al. 1997)

Composition of cement phases

FIGURE 2.3 Reflected light microscopy nest of free lime (magnification 800×).

FIGURE 2.4 Iron-rich octahedra of spinel closely related to magnetite surrounded by interstitial phase (Octahedra 10 γm).

CS are not hydraulic. C3S2 (Rankinite) is also not hydraulic and does not occur in Portland cements. C 3 S and C 2S are the main hydraulic phases, undergoing several phase transformations. The complex solid solutions of C3S and C2S are named ‘alite’ and ‘belite’. Important data on calcium silicates are summarized in Table 2.3.

incorporated in belite. At 1500°C up to 1.4 per cent K2O, 1.4 per cent Na2O and 1.2 per cent Li2O can be taken up, but in commercial clinkers, normally less alkalis are incorporated. The incorporation of 2.2 per cent MgO and 1 per cent A12O3 and 1.1 per cent Fe2O3 at 1550°C was described. ZnO can be incorporated up to 5 per cent, Cr2O3 1.4 per cent and Ga 2 O 3 up to 1.4 per cent. A combined incorporation stabilizes triclinic and monoclinic polymorphs. A combined addition of A12O3 and Fe 2O 3 leads to an influence on each other on incorporation. A fluorine-stabilized C 3S was described by a formula Ca6-0.5xSi2O10-x F.x. The incorporation of small amounts of Fe2+ in C3S under reducing conditions leads to a destabilization and formation of belite and calcium-ferrite, but can also lead to increased free lime contents. A diagram of varying contents of MgO and SO3 showing the stabilization of different M3 and M1 modifications in industrial clinkers was determined by Maki. A zonal structure of alite is obtained due to lower cooling (less foreign ions), and the size of alite crystals is influenced by residence time in the furnace. Due to rapid cooling, often a zonation of alite crystals can be found caused by M 1 and M 3 modifications or varying chemical compositions. The chemical variations and varying reactivities are not well established yet, but it seems to be clear that the inner cores of the crystals contain increased contents of foreign ions. The determination and

2.3.1 C3S—alite C 3S, a highly hydraulic compound occurs in amounts of 50–90 per cent in Portland cements. C3S is formed above 1250°C by a reaction of C2S and C and can be metastably obtained by rapid cooling of the mixtures. The formation of C 3S and its polymorphs is highly influenced by other ions present, so high SO3 contents even can suppress alite-formation which otherwise can be partly avoided by the presence of fluorine-ions. C3S undergoes several phase transformations which are summarized in Figure 2.5. The high temperature polymorph R melts congruently at 2070°C. The polymorphs are all structurally similar with only small distortions, therefore they can be described by pseudohexagonal unit cells. By Rietveld refinement, mainly monoclinic polymorphs can be identified in clinkers. A stabilization of polymorphs can be obtained by incorporation of foreign ions in alite-lattice. Due to foreign ions and stabilization of C 2 S, only small amounts of alkalies are incorporated in C 3 S; instead alkalies are

29

30

*Transformation point. SS=Solid solution; 2H=Ramsdell symbols.

TABLE 2.2 Simple oxides and elements occurring in cements

31

*Phase transition.

TABLE 2.3 CS in cements and their properties

H.Pöllmann

quantification of the C3S modifications by X-ray diffractometry is increasingly simplified by using the structural models and calculations of Rietveld methods. A comparison of X-ray diagrams of high temperature R-phase to monoclinic phase is shown in Figure 2.6. The crystalline structural model is given in Figure 2.9 and the SEM micrographs of single pseudohexagonal shaped C3S is given in Figures 2.7 and 2.8.

FIGURE 2.5 Polymorphic phase transformations of C3S.

The crystalline structure was determined by Maki. A triclinic structure determination was done by Golovastikov (1975). Monoclinic alites show pseudotrigonal symmetry, in clinker often M1 or M3 and mixtures of both in zoned crystals are quite common (Kato 1978; Maki 1982). The structure is composed of SiO4-tetrahedra which are linked by calcium ions. Calcium is coordinated by eight oxygens (1978, 1982).

FIGURE 2.7 SEM micrograph of pseudohexagonal alite crystals showing slight surface coatings from starting hydration.

FIGURE 2.6 Comparison of X-ray powder data (CuKα) diagrams for rhombohedric high temperature C3S and monoclinic C3S. 32

Composition of cement phases

FIGURE 2.10 Polymorphic transformations of Belite. FIGURE 2.8 SEM micrograph of alite crystals separated by interstitial phase.

FIGURE 2.9 Crystal structure of C3S.

2.3.2 C2S—belite C2S mainly occurs in an amount of 10–40 per cent in OPCs. In special cements like belitic cements, low energy cements and slag materials, different modifications of C2S may occur. C2S undergoes several phase transformations in the temperature range (Figure 2.10). The phase transformation to γ-C2S is known as ‘dusting’ due to an enormous volume change. γ-C2S is barely hydraulic. β-C2S can be obtained by quenching or crystallochemical doping. In cements, often α- or β-polymorphs occur. These phase transformations cause the twin lamellae which can be seen in the optical microscope. The different C 2 Smodifications can be stabilized at room temperature with doping by B, P, Mg, alkalis, Ba,

Sr preferentially. Mainly, rather large or small ions have stabilizing effects. It is quite obvious that belite is capable of incorporating larger amounts of foreign ions than alite. The stabilization of highly reactive α-polymorphs by foreign ions using alkalis is used in technical applications. Optically, four different types can be distinguished (F.P. Sorrentino 1998). Activated belite can be obtained by rapid cooling, lattice distortions by solid solutions, crystal size, crystallochemical changes and chemical shift of bonding energies. A comparison of X-ray powder diffractograms of three different polymorphs is shown in Figure 2.11. In OPC clinkers, different belite origins coming from different sources can occur: 1. Primary belite formed by reaction of lime and silicon sources. 2. Secondary belite formed by the decomposition reaction of Alite according C3S ⇒ C2S+C and forming small crystals on the rims of alite crystals. 3. Tertiary belite due to the recrystallization of the interstitial phase coming from the decomposition and decreased solid solution of SiO2 in C3A-phases. Some compositions of belites in different clinkers were given by (F.P.Sorrentino) (1998): OPC white cement

33

H.Pöllmann

FIGURE 2.11 Comparison of X-ray powder data (CuKα) diagrams for different polymorphic phases of C2S.

stabilization of β-C2S by excess CaO and of γ-C2S by excess SiO2 was reported. Bredigite is now known as an one phase with a composition of Ca14Mg(SiO4)8. Bredigite was formerly believed to be similar to α-C 2 S but contrary to belite modifications, it is not hydraulic at room temperatures. The structural model of these phases and their polyhedral links of SiO4-tetrahedra linked by calcium is shown in Figure 2.12

FIGURE 2.12 Crystal structure of C2S.

iron-rich cement oil-well cement

The attempts to form active belitic cements by doping C2S or rapid cooling are increasingly interesting because of lowered temperature requirements and also because of the possibilities of using other lowerquality raw materials for cement production. The 34

1. α-C2S, β-C2S, γ-C2S A comparison of selected X-ray diagrams is summarized in Figure 2.11. The crystal structure models of α-C2S is given in Figure 2.12. Optical and SEM micrographs are given on Figures 2.13–2.16. 2.4 Aluminum silicates, calcium aluminium silicates, their properties and occurrence in cements Aluminum silicates and calcium aluminum silicates do not play an important role in OPC production, but metakaolinite and mullite can occur as decomposition products of clay materials in raw meal composition. Calcium aluminum silicates as gehlenite and anorthite typically occur in

Composition of cement phases

FIGURE 2.13 Optical reflected microscopical ragged belite crystals (magnification 700×).

FIGURE 2.14 Optical reflected microscopical ragged belite crystals (magnification 700×).

FIGURE 2.15 Reflected-light microscopy, belite aggregates (magnification 700×).

FIGURE 2.16 Reflected-light microscopy, belites with typical lamellae structure (magnification 700×).

slags and ashes from different sources. Gehlenite is also a typical mineral for high-alumina cements containing increased contents of silicon. Yoshiokaite is a metastable mineral showing some compositional variation within a nephelite-similar structure. Compositional variations within the calcium aluminum silicates with glassy structure occur in glass cements. Table 2.4 summarizes the most important data on some of the phases.

C3A is the most abundant Al-containing phase in OPC, but CA is the most important for high-alumina cements. Due to the Ca/Al-ratio, also CA2 and CA6 can occur in special HACs. C12A7 is not a pure compound in the system CaO–Al2O3, but instead, must be given by the formula C11A7·Ca(OH)2. Ca(OH)2 can be replaced by CaF2, CaCl2 and others.

2.5 Calcium aluminates

Several polymorphs as a function of Na, Fe and Scontent were synthezised by Lee, F.C., Banda, H.M. and Glasser, F.P. (1982). The cubic pure C3A crystallizes in Pa3, but cubic C3A doped with alkalis seems to crystallize in P213.

In the system CaO–Al2O3 (Rankin and Wright 1915) four binary and one pseudobinary compounds exist (see Table 2.5).

2.5.1 Tricalcium aluminate—C3A

35

36

TABLE 2.5 Calcium aluminates in cements and their properties

TABLE 2.4 Aluminum silicates and calcium alumnium silicates, their properties and occurrence in cements

Composition of cement phases

Different stabilized cubic, orthorhombic and monoclinic lattice parameters are given by Moranville-Regourd and Boikova. The crystal structure is composed of (AlO 4) 5– -tetrahedra linked to (Al6O18)18–-rings which are connected by the Ca2+-ions. For solid solutions with alkalies 2 Na+ replace Ca2+. Therefore a Na+-ion takes the place of the Ca2+ and the second is located in the centre of the (Al6O18)18–-rings. Metastable forms of C 3 A (proto-C 3 A) were synthesized by crystallization of highly undercooled melts with increased contents of Fe2O3 and SiO2 (Lee, Bandd and Glasser 1982). The crystal structure of an orthorhombic complex solid solution with the composition Ca0.8375Na 0.875[Al 5.175Fe 0.45Si 0.375O 18] Was determined by Takeuchi, Nishi and Maki (1980). The orthorhombic and monoclinic solid solutions of C3A can be transformed at high temperatures (T>450°C) into tetragonal (pseudoorthorhombic) polymorphs (Meissner 1986). The Na-doped tricalcium aluminates seem to show decreased reactivity. The ranges of solid solutions and the symmetry changes can be given by: 0–1.9% 1.9–3.7% 3.7–4.6% 4.6–5.9%

Na2O Na2O Na2O Na2O

FIGURE 2.17 Crystal structure of tricalciumaluminate.

cubic cubic+orthorhombic orthorhombic monoclinic

These contents of sodium can be increased significantly in the presence of other foreign ions. The existence of a NC8A3 can probably only be confirmed in the case when Si is present and rapid cooling of melts is occuring. A similar compound containing potassium can also be obtained preferentially when additionally iron and siliconions are present. A tetragonal KC8A3 was synthesized by Adams. C3A crystallizes in a cubic lattice, but due to incorporation of foreign ions, mainly alkalies and SiO2, the symmetry changes to orthorhombic and monoclinic point groups. A detailed structure schemata is given in Figure 2.17. SEM micrographs of tricalciumaluminate are shown in Figures 2.18 and 2.19. X-ray data examples of tricalciumaluminate and sodium containing phases are given in Figure 20.

FIGURE 2.18 Molten calcium aluminate phase—SEM micrograph of molten interstitial phase, mainly aluminate.

FIGURE 2.19 Cubic tricalcium aluminate—SEM micrograph of cubic C3A-crystal edge. 37

H.Pöllmann

FIGURE 2.20 X-ray diagrams of typical cubic, orthorhombic and monoclinic tricalciumaluminate with different sodium contents.

2.5.2 Other calcium aluminates Monocalcium aluminate is the main phase of HAC and crystallizes in a monoclinic point group. The hydroxide in C11A7·Ca(OH)2. can be replaced by F2, Cl2 and others. These phases can play quite an important role in special cements like early hydrating cements. The fixation of halogenide ions is also very important in the combination of burning raw meals with chlorine or fluorine-bearing waste materials. In OPCs, the calcium content is too high, so that these phases cannot be found. Mayenite can also be formed as an additional constituent in high-alumina cements. CA2 and CA6 can occur in aluminumrich high-alumina cements. 2.6 Calcium aluminium ferrites Iron is mainly fixed under oxidizing conditions in the calcium aluminum ferrite phase. The composition of the ferrite phase can be described by a limited solid solution between C2F and C6A2F with C4AxF(1–x), 38

0<x<0.7 (Table 2.6). According to Büssem, the crystal structure of C4AF is composed of layers of (Al,Fe)O6octahedra and (Al,Fe)O4-tetrahedra linked along joint edges. In the open spaces in-between, the calcium atoms are located. A natural mineral was named ‘brownmillerite’. Due to the occupation of iron and aluminum in tetrahedral and octahedral sites within the solid solution, a change of space group from Pcmn to Ibm2 takes place at a composition Ca4Fe4–2x Al2xO10 (x<1.14) (Colville and Geller). Preferentially, aluminum occupies the tetrahedral sites. Mn3+ and Cr3+ also can be incorporated in brownmillerite. C4AMn was synthesized by Puertas, F.; Blanco Varela, M. and Dominguez, R. (1990). The crystal structure is shown in Figure 2.21. According to Mayerhofer, brownmillerite shows preferential orientation in (0k0)direction. With decreasing Al-content of the solid solution, lattice parameter bo mainly decreases. But, due to the increased uptake of silicon and magnesium at elevated temperatures and varying cooling rates, lattice parameters of calcium aluminum ferrites are

39

TABLE 2.6 Calcium ferrites and calcium aluminum ferrates occurring in cements

H.Pöllmann

also affected. In quenched laboratory samples, peakbroadening of X-ray peaks is rather common. The intensities of peaks within the solid solution is influenced by the site occupancies of aluminum and iron (Mayerhofer 1996; Walenta 1997) (Figure 2.22).

FIGURE 2.21 Crystal structure of C 4AF showing tetrahedral and octahedral coordination of (Al,Fe)O6 and (Al,Fe)O4 with Ca-ions.

In typical clinkers, the uptake of magnesium, titanium and silicon takes place, and can be given by the following equation:

with a Si/Ti ratio=3:1 (Harrison et al.) The composition of these ferrite phases typically lies within the range: Ca4Al2Fe1.2Mg0.4Si0.3Ti0.1O10. The composition in iron-rich cements can vary from the composition given above. The influence of varying burning conditions and reducing atmosphere on the aluminum/iron ratio was shown by Sieber et al. Due to increased silicon solution at high temperatures, some exsolution can occur at lower temperatures forming the tertiary belite in the interstitial phase. Titanium forms in high-alumina cements phases like perovskite or C-f-T with similar structure to perovskite but with Fe2+ (Motzet and Pöllmann 1998). OPCs normally have only low titanium contents which may increase when addition of wastes like painting dusts are added in fabrication. Then, TiO2-phases or ilmenite may additionally occur. Reducing conditions can lead to the formation of Fe2+. This leads to increased formation of calcium

FIGURE 2.22 X-ray diagrams of ferrate phase showing different aluminum/iron contents. 40

Composition of cement phases

aluminate and free lime. Fe2+ cannot be incorporated in the ferrate phase. In some cases wüstite or spinels can be identified. 2.7 Influence of minor components (alone or in combinations)—alkalies, MgO, SO3, phosphates, fluorides, borates, chlorine, titanium, heavy metals 2.7.1 Alkalis in cement The amount of alkalis and their fixation in cement phases is very important due to their incorporation in cement phases, and also due to crystallization of new alkali phases. Alkalis can also cause alkaliaggregate reaction in concretes. Alkalis in cement can form in the cooler part of the cement kiln by reaction with SO 3 own alkali-sulphates and contribute mainly to the vapourisation of SO3. The formation of syngenite was also described in lower freshness of cements and was the cause of increased setting times and lumpiness due to aeration of cements (J.Bensted). No Na-analogue of syngenite is known. Other alkali phases are given in Table 2.7 as arcanite, aphtitalite, Ca-langbeinite. Calangbeinite improves the paste viscosity. Alkalis can also be incorporated in cement phases. The change of symmetry and the lowered hydration behaviour of alkali-containing orthorhombic C 3 A was described. Alkalis are mainly incorporated in C2S and stabilize the α and α′ polymorphs. But also the increased alite size, the decrease of viscosity of melt and the increase in belite contents of cements are described. 2.8 Increased MgO content This becomes more and more important due to the fact that less good-quality primary materials are available. Normally MgO forms free periclase or will participate in some solid solutions with tricalcium silicate and the calciumaluminumferrate phases. MgO can also be a constituent of spinelstructure types which occur scarcely. MgO also reduces the viscosity of the interstitial melt. Bredigite and merwinite which were reported also contain

magnesium as a constituent. Taylor summarized some other magnesium-containing phases as diopside, akermanite, etc., in combination with SO3 stabilization of MI and MIII polymorphs of C3S polymorphs. 2.9 Sulphate-containing phases Sulphate bound to calcium is normally bound to anhydrite which can hydrate to bassanite or gypsum in ground clinkers. Anhydrite only occurs for high SO3/alkali-ratios. Yéelimite is the main phase in sulphoaluminate clinkers and shows a quite rapid hydrating behaviour. Yéelimite can also be an important phase in shrinkage-compensating mixtures. Silicosulphate or sulphatespurrite formation is mentioned from reaction zones of refractories and clinkers. Ellestadite is a typical non-hydraulic mineral with apatite structure capable of incorporating different toxic cat- and anions. C2S·CaSO4·CaCl2 is a mineral which is stable up to about 750°C, which is quite hydraulic and can be used in special cement applications. Low burnt sulphate rich waste-mixtures can contain this phase. In Table 2.8, some phases and their properties are summarized. 2.10 Chlorine-containing phases Chlorine contents can be accomodated by raw materials, alternative fuels, combustion of chlorinecontaining wastes or production of low energy cements due to the addition of CaCl2. The alkali chlorides can be vapourized and enriched in kiln dusts. In combustion products high in alkali and aluminum, the phase C11A7·CaCl2 can often be detected. In presence of sulphate chloro-ellestadites can crystallize. Chloro-ellestadites can incorporate different toxic elements, but are not hydraulic. At lower temperatures, C2S·CaCl2·CaSO4, a hydraulic phase form (Kammerers 1993). The use of different kinds of wastes as garbage combustion ash, lime and sulphate can result in the production of alinite-cements with the main component alinite. The composition and diadochic substitutions are summarized in Table 2.9a. Belinite 41

42

Two types exist: alkalis incorporated in clinker minerals and those forming own minerals.

a

TABLE 2.7 Major alkalia phases in cements

43

Other suphates containing additionally alkali-suphates.

a

TABLE 2.8 Sulphate containing phases (without alkalis) in cementsa

H.Pöllmann TABLE 2.9a Alinite compositions with several solid solutions and incorporation of different ions (Von Lampe 1986; Neubauer; Pöllmann 1992)

† Empty lattice space.

was described by Von Lampe, F. et al. Chlorine is also able to form with alkali-elements and heavy metals low-temperature volatile compounds which can lead to special enrichment cycles. Stemmermann and Pöllmann described several other calcium-silico-chlorides phase equilibria and different polymorphs. A natural mineral jasmundite (No. 33–0296) has a alinite-like structure. Different alinites are summarized in Table 2.9a. Various chlorine-containing phases and their properties are listed in Table 2.9b. Figures 2.23 and 2.24 show two typical chlorine containing phases. 2.11 Phosphate in cement In raw materials, normally there are only low P2O5 concentrations, but due to water-clarification muds as a secondary raw material increased phosphate contents occur. The systems CaO–Al2O3–P2O5 and CaO–SiO2–P2O5 were investigated by Hill, D.S., Faust, G.T. and Reynolds, D.S. (1994); St. Pierre, 44

P.D.S. (1956) and Welch and Gutt (1967); Trömel, G (1943). Phosphate only enters tricalcium silicate in very limited concentrations but is reported to form some limited solid solution with C 3 P. At higher temperatures, some complete solid solution may occur. The C4P does not occur in cementitious systems. The reported stabilization of dicalcium silicate tends to stabilize α′-polymorph at lower concentrations, and α-C 2 S at increased concentrations. In the presence of fluorine, the formation of fluoro-apatite was reported. Increased contents of phosphate lead to decreased formation of C3S and increased amounts of C2S-polymorphs (Götz-Neunhöffer and Neubauer 1998). Silicocarnotite and nagelschmidtite are other reported phases. Others with varying stochiometry exist, but have not yet been identified in cementitious systems. The most important phosphorouscontaining phases are listed in Table 2.10.

45

TABLE 2.9b Cl– -containing phases in cements

H.Pöllmann

FIGURE 2.23 SEM micrograph of Sylvite on top of dust material.

2.12 Heavy metals in cement Heavy metals normally occur only in minor quantities in cements and do not cause any problems. Only small amounts of ZnO normally come with the raw meal. An enrichment can take place when car tires or slags from metallurgical processes are added. But, often, artificial addition is made because ZnO improves the burning behaviour and lowers the temperature. The addition of ZnO enlowers the clinkerization temperature (1.5 per cent=50°C and 2–4 per cent=100–150°C). The phase formation with CaO is developed by addition of ZnO. In combination with sulphate and fluorides, burning temperature can be lowered to about 1050°C. According to Odler up to 4.7 per cent can stabilize different polytypes of C3S. Other phases with Zn were described—Ca2ZnSiO5 and different calcium aluminum zinc oxides. Chromium especially can be oxidized to Cr6+, forming calcium chromate. Due to their causing skin eczema, the contents of chromium6+ are strongly limited and reducing agents such as Fe2+-salts are sometimes added. Despite this, the contents are very low and chromium will be incorporated in ettringite on hydrating. Other heavy metals only occur in small quantities as we know from the raw materials, carbonates and clays. Therefore, we normally do not detect single phases of these elements. The formation of low volatile compounds with sulphate 46

FIGURE 2.24 SEM micrograph of crystals of C2S·CaSO4·CaCl2.

or halogenides can cause an enrichment of these phases as was demonstrated with Tl(Cl,Br)-crystals by Bambauer. Addition of different kinds of wastes or alternative fuels can lead to an increase of heavy metals and should be controlled carefully. The use of CuO and Co2O3 were described as fluxes by Odler and Abdul-Maula (1980). Cadmium can replace calcium in different phases. It can replace Ca in tricalcium aluminate, but low volatile cadmium halogenides can be formed. Barium preferentially enters aluminates forming BaO·Al2O3 (Jovanovic 1992). Some selected heavy metal containing phases are given in Table 2.11. 2.13 Elements used as fluxes (boron and fluorine) Fluoride can be present in cements as a fluxing agent or occurs as minor quantities from natural sources where F –—can replace OH –—in micas. Other sources can be some secondary raw materials used in cement manufacturing. Fluorides accelerate the formation of silicates in the burning of the raw meals. Besides 2C2S·CaF2, Ca6–0.5xSi2O10–xFx is also described. Fluoride can also react with the aluminate phase to form calcium aluminium fluoride compounds. The compound C11A7Ca(OH)2 forms a complete solid solution series with C11A7CaF2, a sodalitic compound C3A3CaF2 also exists (Table 2.12). These phases mainly play some

47

Many other stochiometries from different phases exist.

a

TABLE 2.10 Phosphate-containing phases in cements

48

Different slightly volatile halogenides of heavy metals not included.

a

TABLE 2.11 Selected phases from cements containing Zn, Cd, Pb, Ba, Mna

49

TABLE 2.12 Phases occuring in presence of fluxes(B3+, F–)

H.Pöllmann

important role in special cements, slags and metallurgical problems. F– can also be incorporated instead of O2– in calcium silicates. Some A12O3 is probably necessary. Cuspidine Ca(Si 2 O 7 )F 2 does not ocur in cementitious systems. The use of boron in cements is mainly due to the effect of a fluxing agent. To some extent, there is some incorporation in C3S. Due to the amount of addition, different polymorphs of C2S can be stabilized. A calciumsilicoborate is mentioned in Table 2.9. Basic investigations were undertaken by Fletcher. Ellestadite crystallizes as a reservoir mineral capable of fixing different toxic cations and anions (Figure 2.25).

by Bogue calculations but also by direct and fast analysis. By knowing the crystal structures of all phases present in the clinker mixture, it is possible to calculate a theoretical X-ray diagram which will be compared to the measured X-ray diagram of the sample, and a least-squares refinement will take place to adapt both diagrams. By means of

2.14 Quantification of cementitious materials by Rietveld method The important aspect of studying the quantitative phase compositions of mineral phases by Rietveld refinement is possible nowadays. This helps to study the phase compositions of cements not only

FIGURE 2.25 Crystals of fluoro-ellestadite.

FIGURE 2.26 Rietveld-plot with quantification result of a typical Portland cement. 50

Composition of cement phases

this technique, a good and reliable quantification can be installed. Good quantification results were already obtained for Portland cements, high-alumina cements and other construction materials. A diagram illustrating the Rietveld refinement of measured, calculated and difference plot of an OPC is shown in Figure 2.26. 2.15 Acknowledgements All crystal drawings were made using the program Atoms. Many thanks to Prof. Dr. R.Wenda for critical reading of the manuscript. 2.16 References Adams, L.D. (year). ‘Manufacture of Portland cement C3A-compounds, including C3A’, Proc. 14th Int. Conf. Cem. Micr., Costa Mesa, California, 134–44. Abdul-Maula, S. and Odler I. (1992) ‘SO3-rich Portland cements: synthesis and strength development’, Mater. Res. Soc. Symp. Proc., 245, 315–20. Aggarwal, P.S., Gard, J.A. and Glasser, F.P. (1972). ‘Synthesis and properties of dicalcium aluminate, 2CaO·Al2O3’, Cem. Concr. Res., 2, 291–7. Auer, S.T. (1992). ‘Bindung umweltrelevanter Ionen in Ettringit und in Schichtstrukturen vom Typus TCAH’, Dissertation, Erlangen. Bambauer, H.U. and Schäfer, H. (1984). ‘Der Mineralbestand eines thalliumhaltigen Reingasstaubes aus der Zementproduktion’, Fortschr. Min., 62/1, 33–50. Bakdock, P.J., Parker, A. and Sladdin, I. (1970). ‘X-ray powder diffraction data for calcium monoaluminate and calcium dialuminate’, J. Appl. Crystallogr., 3, 188–91. Berggren, J., (1971). ‘Refinement of the crystal structure of dicalcium ferrite, Ca2Fe2O5’, Acta Chem. Scand., 25, 3616–24. Bertaut, E.F., Blum, P. and Sagnieres, A. (1959). ‘Structure du ferrite bicalcique et de la brownmillerite’, Acta Crystallogr., 12, 149–59. Bigare, M., Guinier, A., Mazieres, C., Regourd, M., Yannaquis, N. and Eysel, W. (1967). ‘Polymorphism of tricalcium silicate and its solid solutions’, J. Am. Ceram. Soc., 50, 609–19. Black, P.M. (1969). ‘Rankinite and kilchoanite from Tokatoka, New Zealand’, Miner. Mag., 37, 517–9. Boikova, A.I. (1986). ‘Chemical composition of raw materials as the main factor responsible for the

composition, structure and properties of clinker phases’, 8th ICCC, Rio de Janeiro, Vol. 1, 19–33. Bolio-Arceo, H. and Glasser, F.P. (1998). ‘Zinc oxide in cement clinkering: part 1. Systems CaO-ZnO-Al2O3 and CaO-ZnO-Fe2O3, Adv. Cem. Res., 10, (1), 25– 32. Bowen, N.L. and Greig, J.W. (1924). J. Am. Ceram. Soc., 7, 238–54. Bredig, M.A. (1950). ‘Polymorphism of calcium orthosilicate’, J. Am. Ceram. Soc., 33, 188–92. Bucchi, R. (1981). ‘Role of minor elements in clinker’, World Technol. Part 1, 12(5), 210–31; World Technol. PartII, 12(6), 258–73. Büssem, W. (1937). ‘Die Struktur des Tetracalciumaluminatferrits’, Fortschr. Min., 22, 31. Cervantes-Lee, F. and Glasser, F.P. (1979). ‘Powder diffraction data for compounds in the series Nax(Ca3–xNax)Al2O6’, J. Appl. Crystallogr., 12, 407– 10. Chatterjee, A.K. and Zhmoidin, G.I. (1972). ‘The phase equilibrium diagram of the system CaO-Al2O3CaF2’, J. Mater. Sci., 7 93–7. Chopra, S., Ghose, A. and Young, J.F. ‘Electron-optical studies of stabilized α′- and β-dicalcium silicate’, Abstract. Colville, A.A. and Geller, S. (1971) ‘The crystal structure of brownmillerite, Ca2FeAlO5’, Acta Cryst allogr., 27, 2311–15. Colville, A.A. and Geller, S. (1972). ‘Crystal structure of Ca 2 Fe 1,43 Al 0,57 O 5 and Ca 2 Fe 1,28 Al 0,72 O 5 ’, Acta Crystallogr., 28, 3196–200. Colville, A.A. (1970). ‘The crystal structure of Ca2Fe2O5 and its relation to the nuclear electric field gradient at the iron sites’, Acta Crystallogr., 26, 1469–73. Cruickshank, D.W.J. (1967). ‘Refinements of structures containing bonds between Si, P.S or Cl and O or N.X. beta-Ca2Si04’, Acta Crystallogr., 1, 1948–23; (1964). ACCRA, 17, 685–6. Curien, H., Guillemin, C., Orcel, J. and Sternberg, M. (1956). ‘Hibonite, a new mineral species’, Comptes Rendus hebdomadaires des Séances de l’Académie des Sciences, 242, 2845–7. Czaya, R. (1982). ‘Refinement of the structure of gamma—Ca2SiO4’, Acta Crystallogr. B, 24, 1968– 38. (1971). ACBCA, 27, 848–9. Della Giusta, A., Ottonellog, Seccol (1983). ‘Precision estimates of interatomic distances using site occupancies, ionization potentials and polarizability in Pbnm silicate olivines’, Acta Crystallogr. B, 39; (1990). ASBSD 46, 160–5. Della, M.Roy and Oyefesobi, S.O. (1977). ‘Preparation 51

H.Pöllmann of very reactive Ca2SiO4 powder’, J. Am. Ceram. Soc., 60, 3–4. Dougill, M.W. (1957). ‘Crystal structure of calcium monoaluminate’, Nature (London), 180, 292–3. Eysel, W. and Hahn, T.H. (1970). ‘Polymorphism and solid solution of Ca 3 GeO 5 and Ca 3 SiO 5 ’, Z.Kristallogr., 131, S.40–59. Felsche, J. (1968). ‘The alkali problem in the crystal structure of β-alumina’, Z.Kristallogr., 127, 94–100. Fierens, P. and Tirlocq, J. (1983). ‘Effect of synthesis temperature and cooling conditions of beta-dicalcium silicate on its hydration rate’, Cem. Concr. Res., 13, 41–8. Fischer, R.X. (1995). ‘Die Mehrphasenanalyse in Industrie und Forschung’, 6. Philips-Symposium Röntgenbeugung, Bad Hersfeld. Fletcher, J.G. and Glasser, F.P. (1993) ‘Phase relations in the system CaO-B2O3-SiO2’, J. Mater. Sci., 28, 2677–86. Fletcher, J.G., Borate fluxes in ordinary Portland cement production: a feasibility study’, Ph.D. Thesis, Aberdeen University (1991). Flint, E.P. and Wells, L.S. (1936). ‘The system limeboric oxide-silica’, J. Res. Natl. Bur. Stand., 17, 727–2. Füllmann, T. ‘Quantitative Phasenanalyse von Tonerdezementen’, Dissertation, University of Erlangen 1997. Fumito Nishi and Yoshio Takeuchi (1985). ‘Tricalcium silicate Ca3O[SiO4]: the monoclinic superstructure’, Z.Kristallogr., 172, 297–314. Geller, S., Grant, R.W. and Fullmer, L.D. (1970). ‘LD– magnetic structures in Ca2Fe2-xAlxO5 system’, J. Phas. Chem. Solids, 31, 793. Gebhardt, R.F. (1987). ‘KC 8 A 3: X-ray diffraction characterization of the compound and comparison with NC8A3 and C3A’, Cem., Concr. Aggr., 9(2), 117–28. Ghosh, S.N., Mathur, V.K. and Chopra, S.K. (1984). ‘Low temperature OPC-type cement containing superhydraulic C2S phase’, Cem. Concr. Res., 14, 437–8. Glasser, F.P. (1998). ‘A discussion of the papers ‘Influence of ZnO on clinkerization and properties of VSK Cement’ by D.Bordoli, A.C.Bakuah, P.Barkakati, and P.C.Borthakur, and ‘Hydration of ordinary Portland cements made from raw mix containing transition metal oxides’ by G.Kakali, S.Tsivills and A.Tsialtas’, Cem. Concr. Res., 28(10), 1511–12. Götz-Neunhöffer, F. and Neubauer, J. (1998). ‘Effects of raw meal substitution by sewage sludge on OPC clinker studied by Rietveld analysis’, Porc. ICMA, Guadalajara, 130–8. 52

Golovastikov, N.I., Matveeva, R.G. and Belov, N.V. (1975). ‘Crystal structure of the tricalcium silicate (CaOSiO2)3=C3S’, Kristallografiya KRISA, 20, 721–9. Grant, R.W. (1969). ‘Nuclear electric field gradient at iron sites in Ca2Fe2O5 and Ca2FeAlO5’, J. Chem. Phys., 51, 1156. Grant, R.W., Geller, S., Wiedersich, H., Gonser, U. and Fullmer, L.D. (1968). ‘Spin orientation and magnetic properties of Ca2FeAlO5’, J. Appl. Phys., 39, 1122. Greig, J.W. (1927). Am. J. Sc., 13, 1–14. Guinier, A. and Reyourd, M. ‘Structure of Portland cement minerals’, Proc. 5th Int. Symp. Chem. Cem., Tokio, Vol. 1 pp. 1–32 (1968). Gutt, W., Chatterjee, A.K. and Zhmoidin, G.I. (1970). ‘The join calcium monoaluminate-calcium fluoride’, J. Mater. Sci., 5, 960–3. Gutt, W. (1963). ‘High temperature phase equilibria in the system 2CaO SiO2-3CaO P2O5-CaO’, Nature (London), 197, 142–3. Gutt, W. and Osborne, G.J. (1970). ‘The system CaO–2CaO · SiO2–CaF2, Trans. Br. Ceram. Soc., 69, 125–9. Hansen, L.T., Brownmiller, L.T. and Bogue, R.H. (1927). ‘Studies on the system calcium oxide—aluminaferric Oxide’, J. Am. Chem. Soc., 50, 396–406. Harrison, A.M., Taylor, H.F.W. and Winter, N.B. (1985). ‘Electron-optical analysis of the phases in a Portland cement clinker, with some observations on the calculation of quantitative phase composition’, Cem. Concr. Res., 15, 775–80. Hazen, R.M. (1976). ‘Effect of temperature and pressure on the cell dimension and X-ray temperature factors of periclase’, Am. Miner., 61, 266–71. Hentschel, G. (1964). ‘Mayenit, 12CaO · 7A12O3, und Brownmillerit, 2CaO·(Al,Fe)2O3, zwei neue Minerale in den Kalksteineinschlüssen der Lava des Ettringer Bellerberges’, Neues Jb. Min., 1, 22–9. Hill, W.L., Faust, G.T. and Reynolds, D.S. (1994). Am. J. Sci., 242, 470. Ichikawa, M. and Komukai, Y. (1993). ‘Effect of burning conditions and minor components on the color of Portland cement clinker’, Cem. Concr. Res., 23, 933–8. Ichikawa, M., Ikeda, S. and Komukai, Y. (1994). ‘Effect of cooling rate and Na2O content on the character of the interstitial materials in Portland cement clinker’, Cem. Concr. Res., 24, 1092–96. ICSD: Inorganic Structure Data Base, FIZ Karlsruhe. Ilinetz, A.M., Melinowskiiya and Newskiinn (1985). ‘The crystal structure of the rhombohedral modifikation of tricalcium silicate’, Doklady Akademii Nauk SSSR, DANKA, 281, 332–6.

Composition of cement phases Imlach, J.A. and Glasser, F.P. (1973). ‘Phase relations in the system CaO-Al 2 O 3-FeO-Fe 2O 3 at liquidus temperatures’, Br. Ceram. Trans. J., 72, 221–8. JCPDS: International Centre for Diffraction Data. Jeeveratnam, J., Glasser, F.P. Dent-Glasser, L.S. (1964). ‘Anion substitution and structure of 12CaO · 7A12O3. J. Am. Ceram. Soc., 47, 105–6. Jeffrey, J.W. (1967). ‘The crystal structure of tricalsium silicate’, Acta Crystallogr., 1, 1948–23 (1952). ACCRA, 5, 26–35. Jost, K.H., Ziemer, B. and Seydel, R. (1977). ‘Redetermination of the structure of β-dicalcium silicate’, Acta Crystallogr., 33, 1696–700. Jost, K.H. and Ziemer, B. (1984). Cem. Concr. Res., 14, 177. Jovanovic, I. (1992). ‘Untersuchungen in den ternären Systemen CaO-CdO-Al2O3 und CaO-BaO-Al2O3’, Dipl. Arb. Erlangen. Kakali, G. and Parissakis, G. (1995). ‘Investigation of the effect of Zn oxide on the formation of Portland cement clinker’, Cem. Concr. Res., 25(1), 79–85. Kakali, G. (1998). A reply to the discussion of the paper, ‘Hydration of ordinary Portland cements made from raw mix containing transition element oxides’ by F.P.Glasser’, Cem. Concr. Res., 28(10), 1513. Kammerer, F.: Gleichgenichts reaktionen in System CaOSiO2-CaSO4-CaCl2-Dipl.-Arbeit(1993). Kapràlik, I., Hanic, F.Havlica, J. and Ambrùz, V. (1986). ‘Subsolidus phase relations in the system CaO-Al2O3SiO2-Fe 2O 3-MgO-CaSO4-K 2SO4 at 950°C in air referred to sulphoaluminate cement clinker’, Br. Ceram. Trans. J., 85, 107–10. Klemm, W.A. and Skalny, J. (1977). ‘Selective dissolution of clinker minerals and its applications’, XII Cont. Silikate Industry and Silikate Science, Budapest. Knöfel, D. (1978). ‘Beeinflussung einiger Eigenschaften des Portlandzementklinkers und des Portlandzementes durch ZnO und ZnS’, Zem.-KalkGips, 3, 157–61. Leary, J.K. (1962). ‘New compound in the system CaOAl2O3-CaF2’, Nature, 194, 79–80. Lee, F.C., Banda, H.M. and Glasser, F.P. (1982). ‘Substitution of Na, Fe and Si in tricalcium aluminate and the polymorphism of solid solution’, CCR, 12, 237–46. Lieber, W. (1967). ‘Einfluβ von Zinkoxyd auf das Erstarren und Erhärten von Portlandzementen’, Zem.-KalkGips, 20, 91. Lieber, W. and Gebauer, J. (1969). ‘Einbau von Zink in Calciumsilicathydrate’, Zem.-Kalk-Gips, 22, 161. Mackenzie, K.J.D. and Alasti, H. (1979). ‘Formation

kinetics of portland cement clinker phases, II. Tetracalcium Aluminoferrite’, Trans. Br. Ceram. Soc., 162–67. Majumdar, A.J. (1965). ‘The ferrite phase in cements’, Trans. Br. Ceram. Soc., 64, 105–19. Majumdar, A.J. (1965). ‘The ferrite phase in cements’, Trans. Br. Ceram. Soc., 64, 105–19. Maki, I. and Goto, K. (1982). ‘Factors influencing the phase constitution of alite in Portland cement clinker’ , Cem. Concr. Res., 12, 301–8. Maki, I. and Chromy, S. (1978). ‘Characterisation of the alite phase in Portland cement clinker by microscopy’, Il Cemento, 75, 247–52. Maki, I. and Chromy, S. (1978). ‘Microscopic study on the polymorphism of Ca3SiO4’, Cem. Concr. Res., 8, 407–14. Maki, I., Fukuda, K., Seki, S. and Tanioka, T. (1991). ‘Impurity distribution during crystal growth of alite in Portland cement clinker’, J. Am. Ceram. Soc., 74, 2082–5. Maki, I. (1974). ‘Morphology of the so-called prismatic phase in Portland cement clinker’, Cem. Con. Res., 4, 87–97. Maki, I. (1978). ‘Characterization of the alite phase in Portland cement clinker by microscopy’, Cemento, 3. Maki, I. (1978). ‘Microscopic study on the polymorphism of Ca3SiO5’, Cem. Concr. Res., 8, 407–14. Maki, I. (1986). ‘Relationship of processing parameters to clinker properties; influence of minor components’, Proc. 8th Int. Cong. Chem. Cem., VI(I-2), pp. 34–47. Mayerhofer, W. (1996). ‘Rietveld-analyse der Ferratphase: vom Laborsystem zum techn. Portlandze-ment’, Dipl. Arb. Erlangen. McMurdie, H.F. and Insley, H. (1936). ‘The quaternary system CaO-MgO-2CaO · SiO2-5CaO A12O3’, J. Res. Natl. Bur. Stand., 16, 467–74. Meissner, E. ‘Symmetrieänderungen der Mischkristalle im binären System 8(CaO)-Na2O-3(Al2O3)—3CaO · A1 2 O 3 mit besonderer Berücksichtigung der Hydratationsreaktionen. Dipl-Arbeit-Univ. Erlangen (1986). Midgley, C.M. (1967). ‘The crystal structure of beta dicalcium silicate’, Acta Crystallogr., 1, 1948–23; (1962) ACCRA, 5, 807–12. Mondal, J.W. and Jeffrey, P. (1975). ‘The crystal structure of tricalcium aluminate, Ca3Al2O6’, Acta Crystallogr., 31, 689–97. Moranville-Regourd, M. and Boikova, A.I. (1992). ‘Chemistry, structure, properties and quality of clinker’, 9th ICCC, New Dehli, Vol. 1, pp. 3–45. 53

H.Pöllmann Motzet, H.F. and Pöllmann, H.: Quantitative Phase Analysis of High Alumina Cements, Proc. 20th Int. Conf. Cem. Micr., Guadalajara (1998). Mumme, W.G. (1950). ‘Crystal structure of tricalcium silicate from a Portland cement clinker and its application to quantitative XRD analysis’, Neues Jb. Min., Monatshefte (Band-Jahr); (1995), NJMMA, 145–60. Mumme, W., Cranswickl, Onakoumak, S.B. (1950). ‘Rietveld crystal structure refinement from high temperature neutron powder diffraction data for the polymorphs of dicalcium silicate’, Neues Jb. Min., Abhandlungen (Band-Nr.); (1996) NUMIA, 170(2), 171–38. Mumme, W.G., Hill, R.J., Bushnell-Wyeg, Segniter (1950). ‘Rietveld crystal structure refinements, crystal chemistry and calculated powder diffraction data for the polymorphs of dicalcium silicate and related phases’, Neues Jb. Min., Abhandlungen (Band-Nr.); (1995) NJMIA, 169(1), 35–68. Murat, M. and Sorrentino, F. (1996). ‘Effect of large additions of Cd, Pb, Cr, Zn, to cement raw meal on the composition and the properties of the clinker and the cement’ , Cem. Concr. Res., 26(3), 377–85. Neubauer, J., Sieber, R., Kuzel, H.-J. and Ecker, M. (1996). ‘Investigations on introducing Si and Mg into brownmillerite—A Rietveld refinement’, Cem. Concr. Res., 26, 77–82. Neubauer, J.: Realisierand des Deponie Konzeptes der Inneren Barriere—Digs. Erlungen (1992). Newkirk, T.F. and Thwaite, R.D. (1958). ‘Pseudoternary system calcium oxide-monocalcium aluminatedicalcium ferrite’, J. Res. Natl. Bur. Stand., 61, 233– 45. Newkirk, T.F. and Thwaite, R.D. (1958). ‘Pseudoternary system calcium oxide—monocalcium aluminate (CaO·Al2O3)—dicalcium ferrite (2 CaO · Fe2O3)’, J. Res. Natl. Bur. Stand., 61, 233. Niesel, K. and Thorman, P. (1967). ‘The stability fields of dicalcium silicate modifications’, Tonindustrie Zeitung Keramische Rundschau, 91, 362–69. Nishi, F. and Takeùchi, Y. (1975). ‘The Al6O18 rings of tetrahedra in the structure of Ca8,5NaAl6O18’, Acta Crystallogr. , 31, 1169–73. Nishi, F. and Takeuchi, Y. (1985). ‘Tricalciumsilikate Ca 3O[SiO4]: the monoclinic superstructure’, Z. Kristallogr., 172, 297–314. Nishi, F. and Takeuchi, Y. (1979). ‘The rhombohedral structure of tricalsium silicate at 1200 degrees C’, Z.Kristallogr., 149. ZEKRD, 168, 197–212. Nurse, R.W., Welch, H.H. and Majumdar, A.J. (1965). ‘The 54

CaO-Al2O3 system in a moisture-free atmosphere’, Trans. Br. Ceram. Soc., 64, 409–18. Nurse, R.W., Welch, J.H. and Majumdar, A.J. (1965). ‘The 12CaO · 7A12O3 phase in the CaO-Al2O3 system’, Trans. Br. Ceram. Soc., 64, 323–32. Nurse, R.W. (1954). ‘The dicalcium silicate phase’, Proc. 3rd Int. Symp. Chem. Cem., London, 1952, 56–90. Nurse, R.W., Welch, J.H. and Gutt, W. (1959). J. Chem. Soc. , 1080. O’Daniel, H. and Hellner, B. (1950). ‘Zur Structur von 3 CaO·SiO2’, Neues Jb. Min., Monatshefte (Band-Jahr) (1950). NJMMA, 1950, 108–11 Odler and Wonnemann, R. (1983). ‘Effect of alkalis on Portland cement hydratation I.Alkali oxides incorporated into the cristalline lattice of clinker minerals’, Cem. Concr. Res., 13(4), 477–82. Odler, I. (1990) and Abdul-Maula, S. (1983). ‘Polymorphism and hydration of tricalcium silicate doped with ZnO’ J. Am. Ceram. Soc., 66, 1–4. Odler, I. and Abdul-Maula, S. (1980). ‘Effect of mineralizers on the burning of Portland cement clinker’, Zem.-Kalk-Gips, 66, 132–6, 278–82. Odler, I. (1990) ‘Improving energy efficiency in Portland clinker manufacturing’, in Progress in Cement and Concrete, S.N.Gosh (ed.), Vol. I, Part I, 174–201, ABI Books Pvt Ltd, New Delhi. Perez-Mendez, M., Fayos, J., Howie, R.A., Gard, J.A. and Glasser, F.P. (1985). ‘Calcium fluorosilicates: the Ca6– 0.5xSi2O10Fx—phase’, Cem. Concr. Res., 15, 600–04. Perez-Mendez, M., Howie, R.A. and Glasser, F.P. (1984). ‘Ca 3SiO 5 and its fluorine stabilized aristotype: synthesis, stability and postulated structure of Ca6– 0.5xSi2O10-xFx’ Cem. Concr. Res. 14, 57–63. Phillips, B. and Muan, A. (1958). ‘Phase equilibria in the system CaO—Iron Oxide in Air and at 1 Atm. O2 Pressure’, J. Am. Ceram. Soc., 4(11), 445–54; (1959)42, 414. Pöllmann, H., Rohleder, M., Neubauer, J., Riedmiller, A. and Göske, J. (1998). ‘Quantitative phase analysis of cements, Lime, gypsum, building and construction materials by using a new software for quantification’, Proc. 20th Int. Con. Cem. Micr., Guadalajara, 159–74. Pöllmann, H., Neubauer, J., König, U. and Motzet, H. (1997). ‘Clinker quality control using combined methods—A study using microscopy and X-ray techniques—especially Rietveld method’ , Proc. 19th Int. Con. Cem. Micr., Cincinnati, 195–221. Ponomarev, V.I., Kheiker, D.M. and Belov, N.V. (1971). ‘Crystal structure of calcium dialuminate, CA2. Sov. Crystallogr., 15, 995–8. Rankin, G.A. and Wright, F.E. (1915). Am. J. Sci., 39, 31.

Composition of cement phases Regourd, M. et al. (1968). Proc. 5th Int. Symp. Chem. Cem., Tokyo, 44. Regourd, M. (1964). ‘Determination des réseaux de cristaux microscopiques. Application aux différentes formes du silicate tricalcique’, Bull. Soc. Miner. Crist. Franc. LXXXVII, 241–72. Richartz (1986). ‘Einfluβ des K2O-Gehaltes und des Sulfatisierungsgrades auf Erstarren und Erhärten des Zementes’, Zem.-Kalk-Gips, 12, 678–87. Richartz, B. (1994). ‘Reaktion und Abscheidung von Spurenelementen beim Brennen des Zementklinkers’, Schriftenreihe der Zementindustrie, Heft, 56. Rietveld, H.M.A. (1969). ‘Profile refinement method for nuclear and magnetic structures’, J. Appl. Crystallogr. 2, 65–71. Robson, T.D. (1967). High Alumina Cements and Concretes, John Wiley and Sons, New York. Sahu, S. and Majling, J. (1993). ‘Phase compatibility in the system CaO-SiO2-Al2O3-Fe2O3-SO3 referred to sulphoaluminate belite cement clinker’, Cem. Concr. Res., 23, 1331–9. Sieber, R., Pöllmann H. and Brunner, P. (1996). ‘Investigation on linker microstructure, hydration characteristics and quality coming from varying burning conditions in the kiln using waste fuels’, Proc. 18th Int. Con. Cem. Micr., Houston, 284– 303. Singh, V.K. and Glasser, F.P. (1988). ‘High temperature reversible moisture uptake in calcium aluminate Ca12Al14O33-x(OH)2x’, Ceram. Int., 14(1), 59–62. Smirnoff, G.S., Chatterjee, A.K. and Zhmoidan, G.I. (1973). ‘The phase equilibrium diagram of the ternary sub-system CaO-CaO · Al2O3-11CaO · 7Al2 O3 · CaF2’, J. Mater. Sci., 8, 1278–82. Smith,, D.K., Majumdar, A. and Ordway, F. (1965). ‘The crystal structure of γ-dicalcium silicate’, Acta Crystallogr. , 18, 787–95. Smith, D.K. (1962). ‘Crystallographic changes with the substitution of aluminium for iron in dicalcium Ferrite’, Acta Crystallogr., 15, 1146–52. Smith, D.K., Majumdara, A. and Ordway, F. (1967). ‘The structure of gamma dicalcium silicate’, Acta Crystallogr., 1, 1948–23, (1965). ACCRA, 18, 787–95. Sorrentino, F. and Glasser, F.P. (1976). ‘Phase relations in the system CaO-Al2O3-SiO2-FeO. Part. II The system CaO-Al2O3-Fe2O3-SiO2’, Trans. Br. Ceram. Soc., 75, 95–103. Sorrentino, F.P. (1998). ‘Dicalcium silicate in special cements’, Proc. 20th Int. Conf. Cem. Micr., Guadalajara. Sprung, S. and Schmidt, O. (1980). ‘Structure and

properties of Portland cement clinker doped with zinc oxide’, J. Am. Ceram. Soc., 63(1/2), 13–6. Stemmermann, P. and Pöllmann, H. (1992). ‘The system CaO-SiO2-CaCl2-phase equilibria and polymorphs below 1000°C—An interpretation on garbage combustion ashes’ , Neus. Jb. Min. H. 9, 409–31. St.Pierre, P.D.S. (1956). J. Am. Ceram. Soc., 39, 96. Strunge, J. (1986). Einfluβ der Alkalien und des Schwefels auf die Zementeingenschaften’, Dissertation, PhillipsUniversität Marburg/Lahn. Surana, M.S. and Joski, S.N. (1990). ‘Use of mineralizers and fluxes or improved clinkerization and conservation of energy’, Zem.-Kalk-Gips 43, 43–5. Swayze, M.A. (1946). A report on studies of: 1. Ternary system CaO-C 5 A 3 -C 2F-C 2S. 2. The quaternary system CaO-C 5 A 3 -C 2F-C 2S. 3. The quaternary system as modified by saturation with magnesia. Part II, Am. J. Sci., 244, 65–94. Swayze, M.A. (1946). A report on studies of: 1. The ternary system CaO-C5A3-C2F. 2. The quaternary system CaO-C 5 A 3 -C 2F-C 2S. 3. The quaternary system as modified by 5 per cent magnesia. Part I’, Am. J. Sci., 244, 1–30. Swayze, M.A. (1946). ‘A report on studies of CaO-C5A3C2F’, Am. J. Sci., 244(1), 1–14. Takashima, S. (1958). Rev. 12th Gen. Meet., Cement Association of Japan, Tokyo, pp. 12–13. Takeuchi, Y., Nishi, F. and Maki, I. (1980). 7th Int. Con. Chem. Cem., Paris, Vol. VII, p. 426. Taylor, H.F.W. (1990). Cement Chemistry, pp. 45–50, Academic Press, New York. Taylor, H.F.W. (1990). Cement Chemistry, p. 89, Academic Press, New York. Taylor, H.F.W. (1971). ‘The crystal structure of kilchoanite, Ca6(SiO4)(Si3O10), with some comments on related phases’, Miner. Mag., 38, 26–31. Trömel, G. (1943). Stahl und Risen, 63, 21. Tsuboi, T., Ito, T., Hokinoue, Y. and Matsuzaki, Y. (1992): ‘Die Einflüsse von MgO-SO 3 und ZnO auf die Sinterung von Portland klinker’, Zem.-Kalk-Gips, 9, 426–31. Tsurumi, T., Hiranoy, Katoh, Kamiya, T., and Daimon, M. (1994). ‘Crystal structure and hydration of belite’, Ceram. Trans., 13; Supercond. Ceram. Supercond., CETRE, 40, 19–25. Tumidajski, P.J. and Thomson, M.L. (1994). ‘Influence of cadmium on the hydration of C3A’, Cem. Concr. Res., 24(7) 1359–72. Udagawa, S. and Urabek (1978). ‘Crystal structure of modifications of calcium silicate Ca2SiO4 and their phase transformation’, Semento Hijutsu Nembo, 55

H.Pöllmann SGNEA, 32, 35–38. Udagawa, S., Urabek and Yanot (1977). Crystal structure analysis of alpha-Ca2SiO4. Semento Hijutsu Nempo, SGNEA, 31, 26–9. Udagawa, S., Urabek, Yanot, Natsumem (1979). ‘Studies on the dusting of calcium silicate (Ca2SiO4). The crystal structure of alpha-Ca2SiO4’, Semento Hijursu Nempo, SGNEA, 33, 35–8. Udagawa, S., Urabek, Natsumem, Yanot (1980). ‘Refinement of the crystal structure of gammaCa2SiO4’, Cem. Concr. Res., CCNRA, 10, 139–44. Vaniman, D.T. and Bish, D.L. (1990). ‘Yoshiokaite, a new Ca, Al-silicate mineral from the moon’, Am. Mineral., 75, 676–86. Von Lampe, F. et al. (1986). Cem. Concr. Res. 16, 624. Walenta, G. (1997). ‘Synthesen und Rietveldverfeinerungen der Einzelphasen von Tonerdezementen’, Dissertation University Erlangen. Welch, J.H. and Gutt, W. (1960). ‘Effects of the minor components on the hydraulicity of the Ca silicates’, Proc. 4th Int. Symp. Chem. Cem. Monograph no. 43, 1, pp. 59–68. Welch, J.H. and Gutt, W. (1961). ‘High temperature studies of the system calcium oxidephosphorus pentoxide’, J. Chem. Soc., 874, 4442–6. Welch, J.H. and Gutt, W. (1959). ‘Tricalcium silicate and its stability within the system CaO-SiO2’, J. Am. Ceram. Soc. 42, 11–5. Wenxi, H., Yurong, Y., Guangren, Q. and Yangyoun, Q. (1992). ‘The research on C 2A xF 1-x in iron-rich fluoraluminate cement’, 9th ICCC, New Dehli, Vol. 3, pp. 9–15. Woermann, E., Hahn, T.H. and Eysel. W. (1963).

56

‘Chemische und strukturelle Untersuchungen der Mischkristallbildung von Tricalciumsilikat’, Zem.Kalk-Gips, 9(1) Bericht S.370–5. Woermann, E., Hahn, T.H. and Eysel, W. (1967) ‘Chemische und strukturelle Untersuchungen der Mischkristallbildung von Tricalciumsilikat’, Zem.Kalk-Gips, 9(2) Bericht S.385–91. Woermann, E., Hahn, T.H. and Eysel, W. (1968). ‘Chemische und strukturelle Untersuchungen der Mischkristallbildung von Tricalciumsilikat’, Zem.Kalk-Gips, 6(3). Bericht S.241–51. Woermann, E., Hahn, T.H. and Eysel, W. (1969). ‘Chemische und strukturelle Untersuchungen der Mischkristallbildung von Tricalciumsilikat’, Zem.Kalk-Gips, 5(4) Bericht S.235–41. Woermann, E., Hahn, T.H. and Eysel, W. (1969). ‘Chemische und strukturelle Untersuchungen derMischkristallbildung von Tricalciumsilikat’, Zem.Kalk-Gips, 9(5) Bericht S.414–22. Woermann, E., Hahn, T.H. and Eysel, W. (1979). ‘The substitution of alkalies in tricalcium silicate’, Cem. Concr. Res., 9, 701–11. Woermann, E., Hahn, T.H. and Eysel, W. (1979). ‘The substitution of alkalies in tricalcium silicate’, Cem. Concr. Res., 9, 701–11. Yamaguchi and Takagi (1969). ‘Analysis of Portland cement clinker’, 5th ISCC, Tokyo, Vol. 1, p. 181. Yimin, C., Liping, S., Muzhen, S. and Jun’an, D. (1992). ‘Kinetics of aluminoferrite formation at different conditions’, 9th ICCC, New Delhi, Vol. 3, pp. 262–7. Zhmoidin, G.I. and Chatterjee, A.K. (1972) ‘Phase diagram of the subsystem CaO-CaO.Al2O3-CaF2 Transl.f. Izv.Ak.Nauk.SSSR’, Neorg. Mat., 10, 1846–51.

Chapter three

Hydration of Portland cement E.M.Gartner, J.F.Young, D.A.Damidot and I.Jawed

appeared, mostly due to the application of modern instrumental methods of analysis, but a universally accepted view of hydration still awaits detailed formulation. While we are able to describe the hydration processes in considerable detail now, there are still important issues, which await universal agreement. This chapter attempts to address the debatable issues, while still providing an overarching description of the hydration processes. 3.1 Introduction Since the chemical reactions of Portland cement with water are responsible for the setting and hardening of mortar and concrete, the underlying chemistry must be investigated and described in order to understand these processes. However, this is not an easy task because cement is a complex mixture of numerous compounds. In this chapter, the hydration of C3S, C2S, C3A, and the ferrite phase is first discussed and then the hydration of cement. Interactions between the individual hydrating components, and the relationships between hydration and setting and hardening of concrete are considered. An attempt has been made to briefly identify the role of chemical admixtures and minor components (such as alkalis) in hydration; however, lack of space precludes any consideration of the hydration of blended cements. A short account of the microstructure of hardened cement paste and the structure and properties of the hydration products is included because of their importance in determining hardened paste properties. Our knowledge of the hydration processes has expanded considerably since this chapter first

3.2 Hydration of tricalcium silicate Tricalcium silicate (or alite) is the most important constituent of Portland cement. Since the hydration of alite usually controls the setting and early strength development of Portland cement pastes, mortars and concretes, the hydration of relatively pure C3S has been extensively investigated. It is widely recognized that there are important differences between the hydration kinetics of C3S in water and of alite in a Portland cement—water mixture, primarily due to interactions with the other reacting constituents (see Section 3.3). Nevertheless, these differences can in the main be considered to be perturbations on the C3S-water system, which do not greatly change the fundamental reaction mechanisms. Studies on pure C3S-water mixtures have proved invaluable in delineating and understanding the effects of temperature, dilution, admixtures, and other compounds found in concrete. There continues to be considerable divergence of opinion expressed in the literature concerning the mechanism of C 3S hydration, which has been 57

E.M.Gartner et al.

discussed in many previous reviews [1–6]. Two problems make it difficult to directly compare experimental results from different laboratories: first, different preparations of C3S can vary widely in their reactivity; and second, a complete set of experimental data does not exist for any one C3S sample. Therefore, comparisons of results between separate groups can only be made in a general way with considerable latitude for error and misinterpretation. In addition, studies have been made using widely different water-to-solid ratios (w/s) (see Appendix) which have a large effect on the course of hydration. Nevertheless, we believe that the hydration mechanism can be reasonably well explained by a series of hypotheses, only some of which have been clearly proven. In this section, we attempt to provide a unified theory, drawing attention to specific points of disagreement which need further investigation. 3.2.1 Overview of C 3S hydration The hydration of C3S under ambient conditions can be described by reaction (1) (1) However, the calcium silicate hydrate (C-S-H) is an apparently amorphous phase of variable composition, and hence is usually written as C-SH, which does not imply a specific stoichiometry. When the contact solution is saturated with respect to crystalline CH (portlandite) x=1.7 and n=4. The value of ∆H is estimated for x=1.7 and n=2.7, [7,8] although it should not change much with increasing water content as the additional water is very weakly bound. The enthalpy of dissolution of C3S at infinite dilution is estimated to be–132±10 kJ/mol [9], which is consistent with the observation that the dissolution of CH is slightly exothermic. An initial w/s of about 0.42 is necessary for complete reaction. If more water is added, then there will be an excess of liquid at the end of the reaction, but if w/s exceeds about 280, the amount of CH formed will not be enough to saturate the solution, and [CH] will depend on the value of w/s. 58

Since the value of x decreases with decreasing [CH] [9], the C-S-H formed under these conditions will have x<1.7, and n may also vary. If w/s>2200, the solubility limit of all of the potential solid reaction products will be exceeded, and C3S will dissolve completely. If w/s<0.42 then either some of the C3S will not react, or else the C-S-H formed will have n<4. The results obtained at low w/s apparently depend not only on temperature, but also on the way in which the reaction is conducted, indicating that it is very difficult to establish a true equilibrium in this system. In addition, the C-S-H formed is probably a metastable product in most cases. Crystalline afwillite (C3S2H3) may be the thermodynamically stable product in this system at room temperature [10]; however, it only forms if deliberately nucleated [11,12]. The kinetics of reaction (1) are rather complex. Traditionally the early stages of the reaction have been followed by monitoring the rate at which heat is evolved; preferably by using a pseudoisothermal conduction calorimeter, which can be a very sensitive technique because the hydration of C3S is highly exothermic. The instantaneous thermal power output of such a calorimeter (properly corrected for instrumental factors) is a good measure of the rate of enthalpy change of the reaction, and thus of the rate of C3S hydration. An idealized calorimetric curve is given in Figure 3.1a which has usually been divided into five distinct stages [13,14] but, in the light of recent studies, we have added an initial ‘Stage 0’, as shown in Table 3.1. The choice of the calorimetric curve to represent the reaction is purely arbitrary, other experimental observations such as the rate of release of Ca2+ into solution (drawn schematically in Figure 3.1b) could be used. However, the calorimetric curve has the advantage that it is usually closely related to the actual rate of reaction, and when calibrated on an absolute scale, can be integrated to give the total degree of reaction at any given time. Qualitatively it has the advantages of relative simplicity, sensitivity and accuracy, which have made it the most frequently used method, although the exact shape and position of the peaks will depend on the particular instrument used. However, without the capability

Hydration of Portland cement

FIGURE 3.1 Schematic representation of changes taking place during the hydration of C3S pastes (w/s< 1.0).

of mixing all of the materials inside the calorimeter after they have reached equilibrium temperature, Stages 0 and I cannot be measured. Note that the quality of the mixing obtained in typical laboratory calorimeters varies widely, and can lead to differences in the hydration kinetics. (Also the time-constants of most calorimeters do not allow Stage 0 to be resolved even with in situ mixing.) The duration of each of the stages shown in Figure 3.1 varies widely depending on the reaction conditions as well as the ‘inherent reactivity’ and particle size distribution (psd) of the C3S sample used; for example, Stage 0 perhaps lasts for only a fraction of a second for a typical C3S of average ‘cement’ fineness at the low w/s values typical of concrete, but can represent essentially the whole hydration process (of the order of 1 h) in extremely dilute stirred suspensions. The typical rates of hydration (dα/dt) during all of the stages except Stage 0 are shown in Figure 3.2, plotted against the degree of reaction, (α), rather than against time. The induction period (Stage II) is

represented by the only minimum in the curve. The transition from Stage IV to Stage V is somewhat arbitrary; it can be considered to be represented by the decrease in the rate of decline of dα/dt (occurring for this sample at about 25 per cent hydration or one day’s hydration). The slow steady decline in rate during Stage V continues up to complete hydration, which can take years for a sample with a typical cement-like psd. In this respect, C 3S hydration is very different from that of plaster (calcium sulphate hemihydrate), which shows no decrease in the rate of decline, and ends abruptly with complete hydration after, typically, only about one hour [3]. 3.2.2 The early hydration periods 3.2.2.1 The initial fast reaction

The very early reactions between water and C3S have been the most difficult to elucidate because of the high rates of the processes involved and the difficulty of stopping the hydration rapidly 59

E.M.Gartner et al. TABLE 3.1 Typical sequence of the hydration of C3S at low w/s

enough without creating artefacts. Recent studies [15], have made effective use of the method of dilute suspensions to lengthen the early reaction steps and thus resolve them more clearly. On first contact of C3S with liquid water, the calcium, oxide and silicate ions in the surface layers of the anhydrous phase hydrate and pass rapidly into solution, presumably as simple hydrated (‘aq.’) ionic species, represented here by the symbols Ca2+, OH– and . FIGURE 3.2 Rate versus degree of hydration for a typical C3S sample (23 C, w/s=0.5) [3]. 60

(2)

Hydration of Portland cement

The initial reaction is rapid and highly exothermic, due primarily to the protonation of the extremely basic oxide ions. All of the anhydrous ions, but especially Ca2+, solvate by forming a shell of oriented H2O molecules. Reaction (2) represents congruent dissolution (similar to the dissolution of a simple salt), so that the rate should depend on the specific surface area of the C3S and its density of ‘active sites.’ Thermodynamic calculations [16] indicate that C3S would be extremely water-soluble if no other phases precipitated. An advantage of working with dilute stirred suspensions is that the ions created at the surface of the C3S can be removed rapidly enough to prevent immediate precipitation of any reaction products close to the surface. It is then possible to estimate the very high initial rate of dissolution of C3S. For example, the data shown in Figure 3.3 indicate that, at w/s=20 [SiO2]1 is above 1 mM at the time of the first measurement (<1 min). If this rate were maintained for the whole reaction, it would correspond to complete dissolution of C3S

in only about 4h; but this is certain to be an underestimate since the rate is decreasing even before the first measurement. Damidot and Nonat [9] showed that a C3S sample of 310m2kg–1 Blaine, stirred in pure water at w/s=5000, dissolves completely in about 30min, from which we calculate the initial dissolution rate of this sample to be at least 10 –3 mol s –1 , giving a rate of dissolution per unit surface area of at least 1.4* 10– 5 mol s–1 m–2. Even at this extremely high w/s, the initial dissolution rate is not properly resolved by calorimetry, suggesting that even the bestperforming calorimeters currently available probably have time-constants that are too long to resolve such a rapid reaction. At ‘normal’ w/s values, the initial fast reaction may be almost over even before the C3S powder is completely wetted by mixing with water, and the observed hydration rates are always much lower. This proves that the dissolution of C3S is strongly inhibited at all times after the first rapid dissolution, either by a protective surface hydrate coating, or by poisoning of its dissolution sites. 3.2.2.2 The first deceleration period

FIGURE 3.3 Changes in calcium and silica in solution during early hydration of C3S suspensions.

Under most experimental conditions it is observed that once the mean concentrations of calcium and silica in the solution reach a critical solubility product, defined by Barret et al. [17–20] as the ‘maximum supersolubility,’ nucleation and precipitation of an initial product occurs. In principle, it is this transition that must terminate the initial fast reaction, since the surface of C3S becomes altered in such a way as to be less reactive. The formation by local precipitation of a protective coating of a type of calcium silicate hydrate is not the same as the concept of the ‘superficially hydrolysed’ layer, C3Ssh, proposed by Barret et al. [20,21]. Both hypotheses seem at first sight to differ only in the detailed atomic structure of the surface layer. However, the C3Ssh proposed by Barret et al. is not a true surface phase, but a charge-separated transition state on the surface that is expected to occur if the first step in the hydrolysis is protonation of O2– and , to give OH – and , respectively. Clearly, this 61

E.M.Gartner et al.

results in a protonated C3S surface, with a net positive charge, which must be balanced by the hydroxide ions in the adjacent solution. Although Barret et al. calculated a theoretical solubility curve for this C3Ssh which is appreciably lower than that of anhydrous C3S, it seems questionable to calculate the solubility of a charge-separated phase which cannot be more than one molecular layer thick, and is only one of many possible transition states that might occur during the dissolution process. Furthermore, recent surface charge measurements on C3S in dilute suspensions [22] suggests that it is negative, rather than positive as had been previously reported [23]. We conclude that the initial rapid dissolution is terminated by the formation of a protective coating of a type of CS-H that is probably somewhat different from the C-S-H that forms at later ages [24–26] (which must be more stable). The question of whether changes in the ‘intrinsic reactivity’ of the C3S sample used results in changes in the rate of initial rapid dissolution has not been clearly answered. Fierens and Verhaegen [27] studied the hydration of C3S preparations, which had been subjected to different thermal treatments, and found that the apparent reactivity, based on the rate at which a given level of CH supersaturation could be attained, increased with the intensity of a thermoluminescence peak which corresponded with the probability of escape of trapped electrons. However, their results appear to show that this difference in reactivity was manifested largely during the acceleration period, and not during the initial fast reaction and deceleration, in contrast to earlier studies [28]. These differences may relate to the stability of the protective layer or to the amount of protective hydrate formed during the first two stages of the reaction. If the substrate C3S contains more reactive sites, this could presumably increase the mean rate of thickening of the protective coating, which could, in turn, lead to generation of more nuclei for the final hydrates. During the ‘initial heat peak’ of a typical calorimetric curve (which is a combination of Stages 0 and I), calcium concentrations in solution increase rapidly (Figure 3.3), but silica concentrations are only transitory, reaching 62

maximum values approaching 1.7mM and then decreasing again. The Ca/Si ratio in solution at early ages depends strongly on w/s (Figure 3.4) [29], and on time. At very high w/s, or short times, it approaches the value of three associated with congruent dissolution of C3S. The [Ca2+] profile in the first hour of hydration is extremely sensitive to the specific C3S preparation used; frequently, but not always, a [Ca2+] plateau is observed, especially at high w/s (Figure 3.5) [30]. The surfaces of C3S samples after very short periods of immersion in water have been analysed by ESCA [31–34]. (This high-vacuum technique measures the composition to a depth of only 10– 20Å, but averaged over a large area, while changes in binding energy can also be detected by measuring the energy shift of ESCA peaks.) It was found that the binding energy of some oxygen atoms increases slightly after wetting, and also that even fresh ‘anhydrous’ C3S samples show signs of surface hydration which have to be removed by sputtering under high vacuum. The apparent surface Ca/Si ratio versus hydration time curves (Figure 3.6), show an initial dip within a few seconds, followed by a brief increase (but always to

FIGURE 3.4 Variation of [Ca2+]/[SiO2] ratio in solution after 1 min hydration, as a function of w/s ratio [23,29].

Hydration of Portland cement

FIGURE 3.6 ESCA analysis of the hydrated C3S surface during Stage I hydration.

FIGURE 3.5 Release of calcium into solution in the first hour of hydration for different w/s ratios (numbers on curves) and different C3S samples [23,27,29,30].

a value <3), followed again (after about 1 min) by a slower but steady decrease. Although the initial dip seems to indicate incongruent dissolution from the very start, it could arise from surface contamination (such as traces of silica in the water or removal of calcium during rinsing). The subsequent rise does not occur when very high w/s values (>3000) are used [34], and may also be an artefact of sample preparation, although it could be due to some re-adsorption of calcium by the surface layer, as predicted by the ‘double-layer’ theory [23]. The steady decrease after the first minute can readily be explained on the assumption of a steadily increasing thickness of a low Ca/Si hydrated layer, and is less likely to be due to artefacts of sample preparation. Thus, ESCA data support the concept that a fairly uniform coating of some kind (with Ca/Si<<3) develops within the first 10 min even at w/s=10. Although the C/S ratio of the initially precipitated material was originally assumed by Ménétrier [30] to be 1.0, while Vernet et al. [35] used a value of 1.5, the C/S ratio should vary with

the composition of the solution [18,19]. Direct experimental estimation of the C/S ratio in these early stages is almost impossible to obtain other than by the ESCA measurements, which can only be used to estimate Ca/Si if they represent a continuous C-S-H layer whose thickness exceeds the depth that the ESCA method can sample. In the case of partial surface coverage, the precipitated superficial phase would have a lower Ca/Si than is actually observed, but since it cannot have a ratio of less than zero, the data in Figure 3.6 imply that the surface coverage is at least 30 per cent after about 15min. Parallel SEM studies [34] show the rapid precipitation of a few isolated globular particles which is completed within the first minute. Thereafter, the surface becomes progressively damaged by selective pitting on certain crystal faces [33,36,37]. The initial surface hydrate layer detected by ESCA must presumably be fairly homogeneous and uniform in thickness over the surface, so as not to be visible by scanning electron microscope (SEM). Later, thin flakes or foils of C-S-H are seen to develop and grow rapidly at the surface forming honeycomb-like structures [33,36,37] or an oriented ‘fish scale’ morphology [38]. Similar morphologies have been detected by TEM observations in a ‘wet-cell’, which at 30min show very variable distributions of 63

E.M.Gartner et al.

identifiable concentrations of hydrates over the surface, with a wide range of compositions ranging from almost pure amorphous silica up to C-S-H with Ca/Si=1.2 [39]. Skalny, Young and co-workers [23,30] have proposed a mechanism for the early hydration that does not involve congruent dissolution, and thus essentially omits Stage 0 entirely. Initial dissolution is considered to be incongruent with Ca 2+ and OH – moving rapidly into solution leaving a silica-rich surface layer. Subsequent readsorption of Ca2+ on the now negatively charged surface creates an electrical double layer and gives a positive zeta potential. This helps explain the initial dip in the ESCA curve, and why the surface of C3S appears to be initially free of hydration products, (assuming that the silica-rich surface layer parallels the morphology of the original anhydrous surface.) The silica-rich layer is considered to be a disorganized layer containing ions and possibly some silicate protonated dimers, but since it is also likely to contain some calcium for charge balance it could be considered as a C-S-H of low Ca/Si ratio. Continued incongruent dissolution should increase the thickness of the amorphous layer, which eventually reorganizes into C-S-H possibly via a ‘topotactic’ transformation of the type that had been proposed by Fujii and Kondo [40]. If the concept of incongruent and topotactic formation of a silica-rich layer is maintained, this need not exclude subsequent dissolution precipitation processes [41]. Dent Glasser [42] has suggested that this layer can be a quasi-solid, rich in water, which will subsequently form C-S-H. More recent studies making use of very dilute stirred suspensions have shown that the transition from rapid congruent dissolution to a period of deceleration (PQ in Figure 3.7) [43], during which the calcium and silica concentrations decrease from the curve of maximum supersolubility to a curve for a slightly lower solubility product, referred to as the ‘associated path.’ This can be considered to represent the rapid precipitation of a form of C-S-H from the solution (which at high w/ s acts a significant reservoir of ions). This period essentially marks the end of the free dissolution of 64

FIGURE 3.7 Idealized representation of the dissolutionprecipitation processes in C3S suspensions plotted on a C-S-H diagram, showing the associated path (AP) and curves for C-S-H(SI) and C-S-H (SII) (I and II respectively) [44], and curves for C-S-H(s) and C-S-H (m) (A and B respectively) [26].

C3S, but there is still some disagreement about exactly what happens at this point. The period PQ can be seen to be associated with a rapid deceleration in the rate of hydration. At the end of this brief precipitation, the solution composition follows the ‘associated path’ for a long time, while calcium concentrations increase and silica concentrations decrease continuously. In this respect the ‘associated path’ has all the hallmarks of a solubility curve associated with a C-S-H coating forming on the surface of C3S as proposed by Jennings [26], Barret and Bertrandie [44] propose that the associated path is generated by a kinetic balance between C3S dissolution and the precipitation of ‘normal’ C-S-H (and thus, it must lie between the C-S-H supersolubility curve and the normal C-S-H solubility curve.) In principle, a kinetic balance cannot give a unique curve on the lime-silica solubility diagram; however, the data show considerable scatter about the ‘associated

Hydration of Portland cement

path’. Gartner and Jennings [16] developed a more refined thermodynamic analysis of the C-S-H system and introduced the terminology ‘C-S-H(m)’ to indicate that ‘curve B’ was the solubility curve of a ‘metastable’ form of C-S-H that was probably the protective coating on the C3S surface. The fact that ‘curve B’ was obtained from solution data for preparations still containing some unreacted C3S indicates that the system must be reacting slowly and is not, therefore, at equilibrium, but it is likely that some kind of pseudo-equilibrium can exist between whatever is on the surface of the C3S particles (nominally, ‘C-S-H(m),’) and the solution during this induction period. The relative positions of the different proposed solubility curves and the ‘associated path’ are summarized in Figure 3.7. The concept of protective coatings that are in pseudo-equilibrium with their environment, and which protect an underlying phase of high intrinsic reactivity with the environment, is certainly not restricted to cement systems. 3.2.2.3 The induction period

The induction period in the unretarded C3S-water system is just the minimum in the curve between Stages I and III (see Figure 3.2), and the idealized calorimetric curve shown in Figure 3.1 is in fact only observed when chemical retarders are added to the system. A more typical calorimetric curve for

the early hydration period, given by Brown et al. [45] is shown in Figure 3.8, where it can be seen that there is no true induction period, either in water or in solutions of salts that accelerate the hydration. During this period of low heat evolution, the trends established at the end of Stage I continue. [Ca2+] and [OH–] continue to increase, surpassing the theoretical level of saturation for CH (Figure 3.9). C-S-H (‘Type E’) continues to develop on the C3S surface, but the total degree of hydration by the beginning of the acceleration period is typically only 0.1–1 per cent [6]. A maximum saturation factor (defined as the ionic activity product for the relevant solid, divided by its value at saturation) of well over two with respect to CH is reached before it begins to crystallize at a significant rate. The time at which this maximum is reached is strongly dependent on both the w/s (Figure 3.9) and the reactivity of the C3S sample used. At about the same time as the crystallization of CH begins [30], a new C-S-H morphology (Type 0/Type I) can often be observed by electron microscopy [46]. For moderate values of w/s the correlation between the onset of the acceleration period, the formation of crystalline CH, and the formation of Type 0/Type I C-S-H is often quite close, when due allowance is made for the difficulties in detecting the new phases. This was one of the factors that led Young et al. [30] to conclude that it was the rapid removal of CH from solution

FIGURE 3.8 Rate of heat evolution from hydrating C3S with inorganic salts [45]. 65

E.M.Gartner et al.

FIGURE 3.10 Silicate polymerization during hydration.

FIGURE 3.9 Release of calcium into solution as a function of w/s ratio (indicated for each curve) [29,30].

at that point that drove the renewed acceleration of the reaction. However, the observation that addition of crystalline CH to the mixture frequently had no accelerating effect, and often actually retarded the onset of the acceleration period, [47] is more difficult to explain. It was assumed that the high concentration of silica in the solution at early ages poisoned the growth of CH [23,30], which could not resume until either silica concentrations became much lower or CH supersaturation much higher. The presence of C3S has been shown to inhibit crystallization of portlandite from supersaturated solutions of CH [48]. For many years, the theory of Young et al. [30], based on portlandite crystallization as the driver for the re-acceleration of C3S hydration at the end of the ‘induction’ period, co-existed somewhat uneasily with the alternative theory first proposed by Stein and Stevels [24], that it was the nucleation and growth of a more stable form of C-S-H that caused the acceleration. Jenning’s observations [26] on the lime-silica solubility diagram supports a modified version of Stein and Stevels’ mechanism, and Gartner and Gaidis [3] explained in detail how the conversion of the metastable surface coating, [C-S-H(m)] to a more stable form 66

of C-S-H[C-S-H(s)] by a nucleation and growth process could give rise to the observed kinetics of Stages II and III in C3S hydration. Once the more stable product has nucleated, either within the surface layer or in close proximity, growth of the new phase occurs at the expense of the coating. The coating can apparently reform very rapidly to reach an ‘equilibrium thickness’. Thus, in effect, it is the rate of growth of the stable hydrates that controls the rate of dissolution of the anhydrous C3S. A second important factor in the GartnerGaidis mechanism is the suggestion that the initial number of nuclei of C-S-H(s) is actually proportional to the total amount of C-S-H(m) precipitated during the first deceleration period, which explains why there is no true induction period in the absence of retarders. The importance of the presence of C-S-H(s) nuclei is highlighted by experiments in which an artificially-prepared C-SH was added resulting in a very early transition from the first deceleration period to the acceleration period (see Figure 3.13A). More recently, the mechanism has been probed further using stirred suspensions at high w/s (often with CH solutions of various concentrations replacing water) allowing a greater variation in the times of onset and the duration of the different stages of the reaction. In water at very high w/s, the

Hydration of Portland cement

rate of formation of C-S-H increases well before the onset of rapid CH crystallization, which is detected by its endothermic calorimetric peak [15]. Damidot and Nonat [43,49] explained these results in terms of two different types of C-S-H, termed ‘SI’ and ‘SII’, stable at CH concentrations below and above about 20 mM, respectively. (These types may be the same as those observed by Ramachandran and Grutzek [50] in studies of C3S under controlled pH). The initially-formed C-S-H(SI) becomes metastable, and it is then the nucleation and growth of C-S-H(SII) controls the kinetics. This mechanism is fairly similar to that proposed by Gartner and Gaidis [3], except for the details of the structure and composition of the two types of C-S-H. C-S-H(SI) and C-S-H (SII) have very differently-shaped solubility curves which cross at an invariant point (at about 20 mM [Ca2+] and 0.01 mM [SiO2]) and therefore must have significantly different compositions. NMR spectroscopy showed C-SH(SII) to have a Ca/Si ratio>1.7 and to contain fewer highly-polymerized silicate anions than C-SH(SI) (Ca/Si=1.0–1.5). This is in contrast to earlier NMR studies [51,52], while the Ca/Si ratios are also different from those of the C-S-H(s) and C-SH(m) phases as calculated by Gartner and Jennings [16] based on the two solubility curves observed by Jennings [26]. These differences are still not resolved; the compositions calculated by Gartner and Jennings are very sensitive to both the assumed shapes of the solubility curves and the method of computation of the activities of the various ions in solution. Damidot and Nonat emphasize the need for a high lime concentration in solution, close to that of CH saturation, for the onset of a phase transition in CS-H which destabilizes the initial protective C-SH(SI) and allows it to convert to C-S-H(SII), at the same time as permitting an acceleration period in overall hydration. In this respect, their theory helps form a bridge between the ‘CH nucleation’ and ‘CS-H nucleation’ theories. 3.2.2.4 The acceleration period

At the beginning of this period, typically about 1 h after mixing, well under 1 per cent of the C3S is

hydrated, but, at its end, typically at about 5–10 h, this amount has increased to over 10 per cent (see Figure 3.2). The detailed kinetics of the acceleration period are still a subject of some debate; various kinetic models have been proposed to account for the whole process of hydration, but relatively few deal well with the acceleration period. One of the most popular models is the equation first proposed by Avrami [53] for phase changes in initially homogenous media. (3) where t=time and k is a rate constant. For small values of α, this equation is equivalent to α=ktm. The physical significance of the exponent m depends on the following three factors, according to the relationship (4) In this equation p represents the morphology of the growing phase: (p=1 for growth as needles, 2 for sheets and 3 for isotropic growth; s represents the rate-limiting process: s=1 for pure phase-boundary control and 2 for pure diffusion control; and q relates to the rate of nucleation; q=0 for no nucleation and 1 for continuous nucleation at constant rate). This equation has been successfully applied to Stages II and III, but we find a wide range of values quoted for m in the literature, from below or around 1 [54] to 3 [55,56]. The values obtained clearly depend on the source of C3S used and the time period over which the analysis is applied. If the data are taken too late in the acceleration period, the value of m will decrease due to the onset of diffusion control, while at earlier ages it may include the first deceleration period. Gartner and Gaidis [3] analysed their calorimetric data in the form of a plot of d␣/dt versus ␣ (Figure 3.2), which allows for the early hydration without explicitly re-defining the time axis. It is apparent from Figure 3.2 that the initial part of the acceleration stage on this plot is almost a straight line passing through the origin, approximating an 67

E.M.Gartner et al.

exponential form: (5a) or (5b) Similar growth kinetics have also been observed at early ages in the hydration of calcium sulphate hemihydrate [57]. To fit Avrami’s equation to the same data we must eliminate time, which at low values of α, results in the following form: (6) This equation also comes close to fitting the early-age results shown in Figure 3.2, if m>3. There are relatively few good sources of kinetic data on the very early hydration of C3S, but the calorimetric study of Brown et al. [46] (Figure 3.8a) gives additional data which agree very well with the data of Gartner and Gaidis when reanalysed in this way [58], giving m>3 during the first 4 h. Since nuclei must normally be present at the beginning in order for the process to start, we represent the initial nuclei by an initial value of α, (α0), and integrate (6) to obtain: (7) An exponential (or almost exponential) growth rate is indicative of a branching process, (which continually creates more active growth sites) consistent with an exponent of m>3 in the Avrami equation. A hypothesis capable of explaining this type of growth mechanism and relating it to the structure of the C-S-H found in cement pastes has recently been proposed by Gartner [59]. Not surprisingly, this very high order growth process cannot continue for long and after about 4 h of hydration, the rate of increase of the heat production begins to decrease noticeably. Nevertheless, between 1 and 4h the rate has increased by more than an order of magnitude, indicating a doubling time of under 1 h. The decrease in rate is universally agreed to be due to the onset of some kind of diffusion-controlled process. 68

An alternative non-calorimetric approach to understanding the nucleation and growth processes involved in C3S hydration at early ages has been taken by Gauffinet [38], who used a modification of the dilute stirred suspension approach (w/s~ 50) in which the calcium hydroxide concentration in the liquid phase is initially adjusted to a desired value and then remains fairly constant. The degree of hydration of the C3S could be followed by conductimetric measurement of the increasing CH concentration because the system contained no other components and CH did not precipitate during the period of interest. C 3S hydration profiles were observed to be sigmoidal and to depend significantly on the initial CH concentration chosen (Figure 3.11), indicating a growth process starting from a relatively small number of nuclei formed at the very beginning. At [CH]=16mM (below saturation) it occurs at about 12 per cent hydration, very similar to what is observed by calorimetry at low w/s in Figure 3.2. However, at lower [CH] it occurs earlier and at lower degrees of hydration, while at higher [CH] it occurs later and at higher degrees of hydration, the explanation proposed being that the morphology of the C-S-H formed changes with [CH]. Gauffinet et al. [37] used a computer simulation to allow for the possibility of different aspect ratios in the resulting C-S-H product. They concluded that the apparent early termination of the acceleration period, which was prevalent at low initial CH concentrations, was due to the preferentially laminar growth of the C-S-H parallel to the C3S surface under these conditions, which led to earlier surface coverage and subsequently, the earlier onset of diffusion control. Re-analysis of Gauffinet’s data for the early acceleration stage (up to about 5 per cent hydration) gives an Avrami exponent (m)>3, similar to results obtained by calorimetry at low w/s [58]. This dilute suspension study demonstrated clearly that the overall shape of the kinetic curve is similar for a wide range of CH concentrations; only the actual values of rate versus degree of hydration change (Figure 3.11). Furthermore the kinetics are not very sensitive to [CH] or the precipitation of CH.

Hydration of Portland cement

FIGURE 3.11 Precipitation of C-S-H over time from C 3 S suspensions in lime solutions of different concentrations [36].

3.2.2.5 Summary of hydration mechanisms for the early hydration period

While the number of proposed schemes in the literature are legion they can now be grouped into a few major categories, as summarized in Table 3.2. The most recent work tends to support the concept that it is one form or another of C-S-H which controls the rate throughout the early period, either by its growth or its dissolution. The close correlation between the end of the induction period, the attainment of maximum [Ca2+], and the nucleation of CH [30] does not seem to be true in dilute suspensions, which nevertheless follow very similar hydration kinetics. Delayed growth of CH has been attributed to the adsorption of silicates on the [001] faces [61] and delayed precipitation of portlandite in the presence of soluble silicate has been demonstrated experimentally [23,48]. When C3S is added to saturated portlandite solution immediate precipitation of small crystals occurs [47], but they do not grow larger with time because of silica ‘poisoning’ explaining why CH growth is retarded in the C3S-water system even when CH nuclei are deliberately added. The accelerating influence of carbon dioxide [62] can be explained well by Gartner-Gaidis’ mechanism, since it consumes CH (by the precipitation of calcium

carbonate) at early ages and thus creates more early C-S-H nuclei, just as does dilution in a large access of pure water. Calcium carbonate is itself quite effective as a nucleus for C-S-H [63], which should augment the effect. Thus, the latest evidence tends mainly to support theories based on C-S-H control, although the influence of retarders can still be explained fairly convincingly by their influence on CH nucleation (see below). The concept of one poorlyordered phase (C-S-H) nucleating from another requires an abrupt change in short-range order from the intermediate C-S-H structure. It is clear that the high lime concentration in solution is required for this phase transition in C-S-H to occur. If the fundamental structure of C-S-H is considered to be derived from 1.4 nm tobermorite, then the basic structural unit for both C-S-H and CH is a layer of edge-connected Ca-OH octahedra. (In the case of C-S-H the layer is distorted by the condensation of silicate dimers.) Thus, both hydrates develop from a common nucleus of Ca(OH)2 with adsorbed silicate ions (probably, but not necessarily, dimeric). The higher degree of supersaturation with respect to C-S-H, and the greater variation in local composition that is allowed, means that most nuclei would grow into 69

E.M.Gartner et al. TABLE 3.2 Main theories for the onset and termination of the induction period.

C-S-H, and relatively few into crystalline Ca(OH)2. Such a conceptual view removes the confrontation between ‘CH nucleation’ and ‘C-S-H nucleation’ theories by making both part of a single process. 3.2.3 The later hydration periods It is apparent from the calorimetric curves (Figure 3.8) that the change from the first acceleration period to the second deceleration period is part of a continuous downward trend in d2α/dt2 that actually begins in the middle of the acceleration period. This trend must be due to either a change in the chemistry, which might provoke a reduction in 70

growth rates, or the onset of some kind of ‘diffusion control’ due to precipitation of hydrates around the hydrating C 3S particles. All of the available evidence points to the latter hypothesis as being the more likely. Simple observation by SEM is enough to show that a dense layer of C-S-H forms around the hydrating C3S grains (Figure 3.12a). ‘Wet cell’ TEM studies show that this C-S-H has no definite form in the wet state [46], and this has recently been confirmed by AFM studies under wet conditions (Figure 3.12b) [64]. It is not yet clear exactly which diffusing specie is responsible for limiting the rate of hydration in Stage IV, or what exactly is the form of the diffusion barrier.

Hydration of Portland cement

equation was developed by Jander for diffusion control in solid-state sintering. Several investigators have also satisfactorily fitted kinetic data for the deceleration period to a more general form of this equation: (9) Frequently, x>2, which suggests that the reaction is complicated by the changing thickness of the diffusion barrier or by a changing value of the diffusion coefficient (D), which is contained in k D. Ginstling and Brounshtein [68] modified equation (9) to allow for the changing thickness of the diffusion barrier (e.g. in this case, the C-SH coating) around the C3S grain. Their derivation results in the following equation: (10) Furthermore, C-S-H appears to form with different morphologies at different times and locations, and it is likely that Types I and III C-S-H (see Table 3.3) have different diffusion characteristics from Type IV C-S-H. Taplin [69] was the first to make a kinetic distinction between ‘inner’ and ‘outer’ hydration products, and so derived the following equation on the assumption that the two give different diffusion co-efficients, (Di and Do, respectively): (11) FIGURE 3.12 Growth of C-S-H around C3S particles: (a) SEM micrograph, ×500; (b) atomic force microscopy image (photograph courtesy of Dr. E Gauffinet and Dr. A.Nonat.). 3.2.3.1 The second deceleration period

Several reviews of kinetic models have been published [65,66]. A wide variety of mathematical expressions have been used to fit the rate versus time curves in this period, but equation (8) is the most commonly used: [66,67] (8) where α is the degree of hydration at time t, and kD is the rate constant for the diffusion process. This

where v is the volume ratio of outer to inner product. The differences between Do and Di do not imply differences in chemical composition, since they could be due to changes in physical properties, for example, density or porosity. The rate constants contain terms related to particle size, so that in polydisperse samples, such as are obtained by normal milling procedures, different sized particles may reach a particular ratedetermining situation at different times. The apparent overall kinetics will be a complex function and lead to erroneous conclusions about rate constants and rate mechanisms if corrections are not made. The problem was first pointed out by 71

E.M.Gartner et al. TABLE 3.3 Types of C-S-H morphology observed during the various stages

Taplin [69], but it required extensive computations [66]. The problem can be approached by treating the particle radius (r) as another variable in the rate equation, solving for each particle size obtained from an experimentally determined psd curve, and summing over all sizes:

TABLE 3.4 Diffusion rate constants for two samples of alites [67]

(12) where mi is the mass fraction of particles of initial radius ri, and αi is their corresponding degree of hydration. At any given time the thickness of the CS-H coating will be greater around grains of large ri, and diffusion control will thus be reached at lower values of αi. Knudsen [70] used a psd equation combined with either parabolic or linear kinetics. The exact kinetic equation used is not critical. The question of whether the effects of psd are equally important at all ages is less easily answered. Many studies have tried to combine the whole hydration process into a single set of equations, but this is intrinsically difficult because it is the earlyage reactions that are less ‘localized’ than the later reactions. Brown [71] found that it required as many as 50 different size classes to simulate the influence of psd of a typical ball-milled C3S on the hydration kinetics. His results indicated that the effects of psd were much more important at early ages than at later ages, and that no single expression could account for the kinetics over the whole period of hydration, although the hydration kinetics beyond the end of the acceleration period could be represented by a relatively simple and (apparently) 72

a

A and B are different preparations of monoclinic alite.

psd-indepen-dent form of the Ginstling-Brounshtein equation: (13) It is not clear how an average value of r is estimated for use in equation (13) or how the model allowed for the transition from the acceleration to the deceleration period. Since acceleration-stage kinetics are implicitly surface-area dependent, since phase boundary reactions are involved, the transition time from Stage III to Stage IV is psddependent, even though the subsequent diffusion-controlled reaction is not very sensitive to the original psd. A typical example of the effects of psd on rate constants for two different alites is shown in Table 3.4. 3.2.3.2. The final slow reaction

It is not likely that there is any very fundamental change in mechanism between the second deceleration period and the final slow reaction.

Hydration of Portland cement

During Stage V, C-S-H that has already formed may densify because less empty space is available. Taylor [72] has suggested that the hydration is probably at least partly ‘pseudomorphic’ during this stage, and, indeed, apparent relics of original alite particles (which can presumably be associated with Taplin’s ‘inner product’) are frequently observed in microscopic examinations of hydrated pastes. [71,73] The degree of silicate anion polymerization in the C-S-H increases steadily with time during Stage V (Figure 3.10) which implies that the average C-S-H in the pastes is probably tending toward a lower Ca/Si ratio. It is not known if there is a compositional difference between the two principal C-S-H morphologies (Types III and IV; see Table 3.3) which are observed at this time. Portlandite crystals continue to grow larger, filling large water-filled voids, and can apparently totally engulf zones of C-S-H and even some of the hydrating C3S grains, thereby limiting their potential for complete hydration. These CH crystals always contain small quantities of silica and show quite distinct cleavage on the [001] plane during fracture which may be indicative of the concentration of silica on these planes. Although the kinetics of the diffusioncontrolled period have been modelled in detail, because of its importance in understanding how best to estimate the degree of hydration of real

cement systems as a function of time, the mechanism itself is still poorly understood. Taylor [8,72] has investigated this issue in detail, and has made an interesting comparison between the atomic concentrations of the major elements in alite, belite and C-S-H. In order for C3S to convert pseudomorphically to C-S-H, it must lose about 65 per cent of its original calcium atoms, 40 per cent of its silicon atoms, but only about 6 per cent of its oxygen atoms, while gaining about one proton for every oxygen atom. Clearly, large migrations of Ca and Si must be involved in Stage V hydration and it is reasonable to assume that the rate-limiting process in Stage V must involve diffusion of one of these ions through the hydrated coating. Calcium may be able to diffuse as Ca2+, but silicon is most likely to diffuse as the larger partially protonated orthosilicate monomer. There is evidence that Ca2+ is the more mobile of the two species. The main difference in Stage V is that there is no longer much empty pore space available around each hydrating grain, due to the overlap of the hydrated coatings of neighbouring grains, thereby resulting in a rapid increase in the mean diffusion distance that an atom has to travel between a water-filled pore in which fresh C-S-H and CH can still deposit, and the anhydrous surface of a C3S particle from where the Ca and Si must originate.

FIGURE 3.13 (A) Effect of nucleating phase on the heat evolution of hydrating C3S: (a) unmodified curve; (b)+CS-H; (c)+CaCl2; (d) both added [3]. (B) Effect of set-modifying admixtures on the heat evolution of hydrating C3S: (a) set accelerator (e.g. CaCl2 or high temperature); (b) hardening accelerator (e.g. NaCl); (c) unmodified curve; (d) set retarder (or low temperature). 73

E.M.Gartner et al.

3.2.4 Types of C-S-H formed during the different hydration periods 3.2.4.1 Morphologies

In the course of hydration, several different morphological types of C-S-H have been identified. Various descriptive adjectives have been used to describe the morphologies observed, which depend on the technique used. Two major classifications have been proposed based on SEM [74] and TEM [46] studies; they are compared in Table 3.4, and have already occasionally been referred to in the preceding discussions. There are no demonstrated compositional or structural differences to support either classification, and both schemes must be considered arbitrary. There is clearly a need for a better understanding of the differences between the types of C-S-H formed at different times and locations within the microstructure, based less on morphology than on true atomic structure and nanostructure. The problem with conventional SEM and TEM techniques is that they are very sensitive to the methods used to prepare the sample, which involve drying to very low relative humidities, and consequent deformation of the microstructure. With TEM, at high magnification, there is also a serious risk of significant damage due to the electron beam while the specimen is under observation. On drying or fracturing, the dense C-S-H layer around hydrating C3S particles usually splits and consolidates into splines or needles (Types 1/1′) due to crumpling and rolling up of sheet-like material. The saturated material extends into and fills the space between hydrating grains. 3.2.4.2 Silicate polymerization

Since the first edition of this book, new techniques such as solid state NMR spectroscopy, have provided a wealth of information about the atomic structures and nanostructures developed during hydration. It is not possible to discuss this area in detail, but the highlights are reviewed and reference 8 can be consulted for more information. The use of 29Si CP-MAS NMR spectroscopy has 74

revealed that a hydrated orthosilicate (monomer) forms in cement pastes during the induction period [51,52] and persists in small quantities (~2 per cent of Si) for at least one month. During Stage III C-S-H contains primarily dimeric silicates. Later (1–3 days or 20–30 per cent hydration) dimers polymerize to give pentamers, octomers, etc. [51,52,75]. This process can be modelled as a progressive insertion of bridging tetrahedra into the gaps between dimers to form the ‘dreierketten’ structure of tobermorite. Similar insertion has been observed with alumina [76]. This model suggests that in the initial formation of C-S-H, a dimeric silicate species is being laid down on a Ca(OH)2 layer template. The process of gradual polymerization (Figure 3.12) is an indication of slow chemical change in C-S-H, which is concomitant with, but independent, of its formation by hydration. Rather different results were obtained on examining C-S-H formed by hydration of C3S in dilute suspensions; C-S-H(SI) which formed first was more highly polymerized [43,49]. Since this forms when [CH] < 20 mM it is possible that this form has a radically different structure. Regardless of the model adopted, a picture has emerged which supports the concept of the initial C-S-H which forms the protective hydrate layer having a different local structure from the C-S-H that forms after the induction period. Such a viewpoint helps support the concept of the nucleation of stable C-S-H discussed earlier. It could also explain, why, under the right conditions afwillite might nucleate as the dominant hydration product in the system. 3.2.5 Effects of temperature Temperature has a marked effect on Stages II and III of the hydration periods of C3S. The induction period is shortened as temperature increases, and the rate of heat evolution increases during Stage III. Hydration can occur at temperatures as low as –20°C (if liquid water is present, e.g. in salt solutions), albeit slowly. Measurements have been made of the apparent activation energy based on degree of hydration-time

Hydration of Portland cement

curves. Values obtained during Stages III and IV are typically close to 40 kJ/mol, whereas during Stage V the apparent activation energy approaches 20 kJ/ mol, which is close to the value expected for a diffusion-controlled reaction. It should be remembered that the true activation energy of a reaction can only be measured by means of an ‘Arrhenius’ plot if a single, clearly-defined reaction step is the only process influencing the change of rate with temperature; a situation seldom encountered in cement chemistry. The effects of temperatures significantly removed from ambient condition on C-S-H properties have not been as extensively studied. Morphologies do not appear to differ extensively, but the degree of silicate polymerization increases with temperature. Up to about 80°C, C-S-H appears to have essentially the same properties as that formed at room temperature. However, there are some indications that the Ca/Si ratios of C-S-H change with temperature [77–80]. 3.2.6 Influence of admixtures Admixtures that have a very marked influence on the hydration kinetics can be used to probe the mechanism of hydration. Such admixtures are generally classified as either ‘accelerators’ or ‘retarders,’ depending on how they affect the time of set. The effects of certain ‘typical’ admixtures, (and of temperature,) are shown schematically in Figure 3.13B. However, it is important, (especially in the case of Portland cement,) to note that admixtures that modify the hydration of pure C3S will not necessarily show the same effect when used with portland cement, and vice versa. Furthermore, many admixtures may change their behaviour completely as a function of dosage. There are at least two major classes of accelerator: soluble salts and fine powders, which appear to work by completely different mechanisms. Fine particles that accelerate C3S hydration do so by nucleation of C-S-H. Silica fume and laboratory preparations of C-S-H can effectively shorten the induction period without changing the maximum rate of hydration. On the other hand, accelerators based on soluble salts generally increase the

maximum rate without accelerating the onset of Stage III to any great extent. Because the two mechanisms operate on different steps in the process, their effects are almost completely separable, as can clearly be seen from the results in Figure 3.13B. Although many soluble salts act to accelerate the hydration of C3S, especially during Stages I, II and/ or III, there are many exceptions, for example, soluble phosphates, fluorides, and salts of Zn, Sn and Pb, which are retarders. On the other hand, almost all organic compounds act as retarders. The most thorough survey to date of the effects was made by Kantro [81] who studied the heat liberation rates of C3S, hydrated in different salt solutions of equal normality. His results confirmed that the anion and the cation influence the hydration rate differently. For the chlorides of alkali and alkaline earth metals at a dosage equivalent to 2 per cent CaCl2 · 2H2O by mass of C3S, the order of effectiveness as an accelerator of Stage III at 27°C was influenced by the cation as follows: (14) denotes a quaternary ammonium ion where and H2O means that no additive was used. Clearly, these results indicate that calcium occupies a special position in the ranking, which is, perhaps, not surprising. Apart from that, the data suggest that both the charge and radius of the ions are important, those with high charge and small size being the most effective. A roughly similar trend is observed for the anions, which also make a very important contribution. For highly soluble salts of calcium at the same equivalent concentration, the order of effectiveness was influenced by the anion as follows: (15) Kantro’s data also indicate that alkali sulphates are very effective accelerators, and others have reported on the efficacy of calcium thiosulphate (which is much more soluble than calcium sulphate). These observations imply that divalent anions which do not precipitate calcium can be 75

E.M.Gartner et al.

very effective catalysts, and support the idea that an electrostatic shielding effect may be important, although the mechanism is by no means clearly understood. Figure 3.8 suggests that it is the rate of increase in the rate of heat liberation during the early part of Stage III, (roughly equivalent to d2α/ dt2), that is influenced by soluble calcium salts. Fierens et al. [56] found that the rate constant determined from Avrami kinetics with m=3 (equation 8) is linearly dependent on CaCl 2 concentration up to about 0.6 M. Hence CaCl2 must be influencing the growth of C-S-H, which is assumed to be the rate-determining step for Stages II and III. This is in contrast to earlier work of Kondo et al. [77] who correlated the overall influence of different salts with ionic mobilities and diffusion characteristics, and concluded that acceleration is determined by the rate of ionic diffusion, the non-participating electrolytes simply providing ions for counter-diffusion. However, a chemically-controlled surface reaction seems inherently more likely to explain the rather complicated influence of different salts during Stage III [6]; and it seems likely that the incorporation of Ca2+ into the growing hydrate is part of the rate-determining step, which would account for the strong sensitivity to [Ca2+]. The presence of a mobile anion, however, clearly plays an important role, which is difficult to understand unless related to charge shielding. If, for example, the attachment of a single calcium ion to a growth site on the C-S-H is the rate-determining step, it is apparent that the surface charge will increase by +2 during this step. In order to minimize the free energy of a hypothetical positively-charged transition state, it is presumably necessary to provide sufficiently mobile anions in the close vicinity (mobile in the sense that they can arrive and depart easily). Chloride and bromide are the smallest and most mobile of the monovalent anions in the list above. Both soluble calcium salts and alkali hydroxides can influence the initial dissolution of C3S because of the common ion effect. It should be remembered that crystallization of CH is controlled by the solubility product KCH which is affected by the activities of the two ions involved, as represented 76

by (Ca2+) and (OH–): (16) If we apply the dilute solution approximation (activity~concentration) and we start with a solution of either a soluble calcium salt or an alkali hydroxide, it can be shown that less CH needs to dissolve (either from free lime or from C3S) to saturate the solution relative to CH than would be the case in if we had started with pure water. Thus, although the apparent rate of solubility of the CH from the C3S often appears to decrease, [30,82] high concentrations of a soluble calcium salt or alkali hydroxide in the mix water can allow the system to reach saturation with respect to CH earlier, after a smaller amount of C3S has dissolved. It seems that a high alkali hydroxide concentration can give significant acceleration [83], although 0.04M NaOH has no accelerating effect [45]. Alkali salts that precipitate insoluble calcium salts will generally increase the early rate of dissolution of C 3 S by keeping the calcium concentration low, as has been demonstrated for sodium carbonate and sodium fluoride [30]. As a consequence of the hydroxide and silicate ions which dissolve along with the calcium, the pH increases and a higher concentration of dissolved silica may also be observed, which should precipitate as C-SH once the calcium concentrations begin to rise again. The length of the induction period appears to depend on how rapidly the calcium concentration rises to reach the maximum CH supersaturation, supporting the idea that a certain minimum CH concentration, probably fairly close to CH saturation, is required for the onset of the true Stage III. The more the calcium consumed by early precipitation, the more ‘early’ C-S-H with a low Ca/Si ratio is likely to form, and this may act as a thickened protective barrier around the C3S. Similar considerations apply when C3S dissolves in solutions of metallic salts from which insoluble hydroxides are precipitated. Additionally, if the metal hydroxide can react further with hydroxyl ions to form a complex oxyanion, which may precipitate as an insoluble calcium salt, then rapid hydration will be retarded until these reactions are complete because both [Ca2+] and [OH–] remain low for a long time.

Hydration of Portland cement

FIGURE 3.14 Effect of metal salts which form reactive hydroxides on the heat evolution of hydrating C3S. The first peak is associated with the precipitation and reaction of metal hydroxides, while the second is the main peak of C3S hydration. [After Ref. 81, with permission.]

This was first recognized in the case of zinc salts (equation 17), but is a universal phenomenon (Figure 3.14): (17) Retardation of this kind is evidence that a certain level of Kp(CH) is required in the solution in order for the stable form of C-S-H to nucleate and grow at a reasonable rate. There is, additionally, the possibility of forming insoluble coatings around the C3S grains that can continue to retard the rate even during the acceleration period [84]. The effects of adding acids are somewhat similar to those of adding metal salts that precipitate insoluble hydroxides. In both cases, hydroxide ions are consumed and the level of KCH is depressed. However, in the case of acids that give soluble calcium salts, no precipitates are formed, and the salt itself is generally an accelerator, so little, if any, retardation is observed. Additionally, the excess silica precipitated during Stage I can provide a large number of nuclei for the growth of C-S-H in Stage III. In the case of acids that give insoluble calcium salts the same rules apply [85], with the complication that the precipitating calcium salt may also be a retarder for portlandite or C-S-H growth. This double effect can often lead to very strong retardation. The ability to retard seems to be strongly linked to the formation of calcium complexes. Thus, soluble calcium salts of simple sulphonic acids, which do not complex calcium to any great extent, are

generally non-retarding, although they are not particularly good accelerators. The (soluble) calcium salts of simple carboxylic acids (acetic, propionic, etc.) which form slightly stronger calcium complexes than sulphonic acids, are either weak accelerators or weak retarders, the exception being calcium formate, which is a moderately good accelerator. However, as soon as the complexity of the anion is increased by the addition of a second group (either carboxylic or hydroxyl) that can interact with calcium, the tendency lies strongly in favour of retardation. It has been observed that many organic acids are capable of inhibiting the growth of CH, apparently due to adsorption onto the [001] face, thus preventing normal crystal growth in the same way as postulated for silica poisoning [61]. Many compounds that strongly chelate calcium also have strong retarding capabilities. The strong retardation caused by many organic acids containing the ␣-hydroxy-carboxylate grouping, such as citric, tartaric and gluconic acids, have been attributed to the formation of a strong six-membered chelate ring with calcium:

(18)

The size and number of chelate rings that can be formed by an organic molecule apparently affect the retarding power but electronic effects are also important for chelating ability. Most inorganic ions do not have this ability for strong chelation. The formation of a chelate complex, such as shown in reaction (18) above, typically occurs in solution, reducing the activity of the calcium ion; however, this alone is not enough to cause severe retardation; rather, it has a similar effect to the addition of an acid, that is, it requires that more calcium dissolves in order to saturate the solution with respect to CH, as shown for the case of a neutral chelating ligand (L) below: (19)

77

E.M.Gartner et al.

where 0<x<1, and x<<1 for a strong ligand, that is, almost all of the calcium is bound to the ligand if there is an excess of ligand available. Since the complex ion L.Ca2+ does not behave as Ca2+, more CH must be released by the source, (in this case, C3S) to combine with the free ligand in order to reach any given value of the solubility product. This can give rise to retardation of the type discussed above for acids or compounds that precipitate calcium. However, the retarding effect is frequently multiplied by the ability of certain chelated calcium ions to adsorb onto the growth sites normally occupied by free calcium ions, thereby ‘poisoning’ these growth sites. Evidence for such strong selective retardation comes from the effectiveness of many other organic retarders, (such as many sugars), which do not contain the βhydroxy-carboxylate grouping, and do not necessarily form very strong complexes with calcium in solution. There have been relatively few studies of the solution chemistry in retarded C 3S systems. Jennings et al. [84] observed that the concentrations of free Ca2+ (corrected for complexation) and total silica in solution during C3S hydration at early ages, in the presence of three different retarders (at 2 per cent concentration in the water, and w/s=2,) always fell on or close to the C-S-H(m) solubility curve that was believed to represent the surface coating on the C3S. The data obtained with the strong tridentate ligand EDTA showed lower freecalcium concentrations as expected for strong complexation. The data suggests that calcium chelation may not be necessary per se for retardation, but that the retarder may stabilize the initial C-S-H coating against conversion to the more stable final C-S-H. The fact that free-calcium concentrations can remain low for a long time in such retarded systems implies that the protective layer reaches a fairly stable thickness, beyond which it allows very little extra calcium into solution. Moreover, an increase in the amount of free calcium in solution at early ages may not lead to reduced retardation. Sucrose retards much longer than EDTA, despite maintaining free [Ca2+] close to CH saturation. This may also imply that the C-S-H coating is itself stabilized directly by 78

adsorption of the retarder. These data support the concept that retardation is caused mainly by inhibition of the growth of the more stable type of C-S-H that forms during Stage III. The concept of C-S-H and CH sharing common nuclei, discussed earlier, may also be the key to reconciling the behaviour of retarding admixtures. 3.3 Hydration of dicalcium silicate Dicalcium silicate (or belite) is less extensively studied than C3S because its hydration contributes mostly to long term properties of Portland cement pastes, mortars and concrete. The β polymorph has been the most extensively studied since it is most commonly found in Portland cements. The much slower reactions of β-C2S lead to very slow strength development and for this reason cements formulated to be rich in belites usually contain small quantities – of alite or calcium aluminates, such as, C3A4S, to provide rapid setting and early strength development. The slower rate of hydration of C2S will make its early hydration more amendable to study than that of C3S. The large differences in reactivity between C3S and C2S can be ascribed to the presence of O2– ions in the structure of the former, in addition to SiO44 –. If the initial attack is by water molecules acting as Bronsted acids [2], then, the higher basicity of O2– is clearly driving the process. However, other factors must be operating to explain the distinct differences in reactivity between different polymorphs of C2S, as discussed more fully in Section 3.2. For example, Jost and Ziemer [87] argue that the sharing faces of CaOx polyhedra is an important feature controlling reactivity. 3.3.1 Overview of hydration The reaction of dicalcium silicate is described by equation (20), which is analagous to equation (1): (20) The composition of C-S-H is similar to that formed from C3S. Hydration and the development of microstructure proceed considerably more slowly than for C3S, when both are prepared by solid-state

Hydration of Portland cement

sintering at 1400–1450°C (i.e. normal clinkering conditions). The hydration processes of β -C 2 S show considerable similarities with that of C 3S, but monitoring hydration by pseudo-isothermal conduction calorimetry is more difficult because of the lower enthalpy and slower kinetics of hydration. Nevertheless, the same profile of heat evolution observed for C3S occurs also for β-C2S [29], and it may be appropriate to divide the reaction into the same six stages. The rate of release of heat in Stage 0 has been reported [88] to be similar to C3S, suggesting that this process involved only heat of wetting. But a higher rate of exothermic dissolution of C3S might be expected due to the greater basicity of O 2–, as discussed above. Immediately upon contact with water non-uniform etching is observed and grain boundaries are revealed [89]. Maycock et al. [28] studied the early heat evolution of both C2S and C3S, but unfortunately do not provide direct quantitative comparisons that might provide the answer to this question. Surprisingly there appear to be few studies of the early release of ions into solution, in stark contrast to the extensive studies of C3S. Barret and Bertrandie [44] have shown that rapid release of calcium and silicate ions into solution also occurs with β-C2S, but [Ca2+] is lower than in the case of C3S and increases more slowly. However, the same problems that bedevil the study of C3S also occur here, namely, that different investigators work with C2S preparations that may have different reactivities, using different experimental techniques. Therefore direct comparisons are not always easy to make and it is difficult to assess the validity of apparently contradictory conclusions. A particular problem concerns the stabilization of β-C2S with respect to γ-C2S. The most common means to prepare stable β-C2S in the laboratory is the use of B2O3 but in cements, stabilization is much more complex (see Section 3.3.2). The way the β -polymorph is stabilized can have a significant influence on its reactivity. The first deceleration (Stage I) can also be attributed to the formation of a protective layer of a

metastable form of C-S-H. This conclusion is based on ESCA studies [31,88,90], SEM studies [89] and solution analysis [44]. While initial precipitation of C-S-H may occur as early as 15 s, subsequent growth is slower than in the case of C3S. The C-S-H at this early stage appears to contain hydrated monomeric silica (Q°) [91,92], so it seems reasonable to assume that the nature of the protective coating is similar to that formed by C3S. ESCA studies support this conclusion. The induction period (Stage III) appears to be extremely variable in length with extreme values reported from 9h [90] to over 20 days [92]. Since the second heat peak is very weak and difficult to measure, factors influencing the length of the induction period have not been probed. During Stage III there is a steady increase in [Ca2+] and [OH–] in solution [93], but the rate of increase is much slower than is observed for C 3S. The maximum CH concentration coincides with the onset of Stage III. A large degree of supersaturation with respect to Ca(OH)2 is not observed and this is consistent with the slow hydration rate and the observation that relatively few portlandite crystals form; but those that are seen are generally very large [94]. To a first approximation, the growth, morphology and composition of C-S-H during subsequent hydration (Stages III–V) are not very different from what is observed in C3S pastes, but there are some interesting differences [94]. Type II morphology is seen well beyond Stage II, whereas this form is not seen in C 3S pastes after the induction period unless admixtures are used [95]. The implications of this observation are not clear, but do perhaps point to the importance of [Ca2+] in solution when C-S-H is precipitated as proposed by Damidot and Nonat [43,49]. Second, hydration shells have been observed to pull away from the unhydrated core during fracture [94], which is a phenomenon seldom observed in C 3S pastes, although common in Portland cement pastes. However, the formation of ‘hollow shells’ in the latter case is believed to be due to the formation of AFt networks, as discussed in Section 3.7.1. The CS-H formed in the later stages of hydration has a much more variable morphology than seen in comparable C3S pastes [94]. 79

E.M.Gartner et al.

3.3.2 Effects of polymorphs Whereas the rates of hydration of the different modifications of C3S stabilized by various impurities are not radically different, the same is not true of C 2 S. At room temperature, γ -C 2 S is the thermodynamically stable form, but it is only weakly hydraulic. The other polymorphs are metastable at room temperature and must be stabilized with impurity oxides. While the mechanism of stabilization has been implicitly attributed to the effects of accommodating the impurity ions within the crystal lattice, the situation is more complicated. Recently it has been shown that other phases rich in impurity ions, and which are partially amorphous, can be formed at grain boundaries [96,97]. It has been suggested [98] that these phases can provide a physical component to stabilization of the martensitic β→γ transformation, as well as limiting the size of β crystals to below a critical size. The rate of cooling is therefore important [99] since it can control the effects of intermediate transformation, the amount and distribution of second phases and the size of the crystals. For a given stabilizer, α′-C2S (which has a higher level of impurities) has a greater hydraulic activity than does β-C2S, but the actual degree of reactivity depends on the kind of stabilizer used, so that some β-forms are more reactive than some α′-forms [100]. Some kinetic data illustrate these points (Table 3.5). The variable reactivities of these polymorphs have been attributed to differences in ionicity of M-O bonds, (where M is a cation) or to different levels of crystal imperfection including, perhaps, twinning [87] and incoherent precipitates [96]. Highly reactive β-C2S has been obtained using special low-temperature

syntheses (see Section 3.3.4). The reactivity of αC2S in relation to the other polymorphs is not well established; its reactivity may lie between α′ and β. The stabilization of this polymorph, which occurs only at the highest temperatures, is not as easy as it is for β and α′. While γ-C2S is generally described as being nonhydraulic it does show very weak hydraulicity [102,103], about 25 per cent hydration being reported after 5 years [102], or 40 months [103]. Trettin et al. [88] reported that γ-C2S is hydraulically active when first mixed with water, but that the protective coating of CSH formed early on inhibits subsequent hydration. Thermodynamic calculations suggest that the free energy of hydration of γ-C2S to give C-SH(m) plus CH is very small [16,44]. However, γC2S is also highly reactive towards carbonation [104]. The reason for the differences in reactivity between the different polymorphs is still a matter for debate. Jost and Ziemer [87] emphasize the importance of face-sharing CaOx polyhedra and the mean Ca-Ca distances in the crystal. Both α′ and β have distorted CaO 7 polyhedra, while γ has a relatively symmetrical closed-packed CaO6 octahedra which do not share faces. Thus the symmetry and strength of the primary coordination shell of Ca is clearly important. However, as pointed out earlier, the complex microstructures that develop during phase transformation must play a significant role. 3.3.3 Reactive dicalcium silicate Over the years there has been a continuing interest in the development of belite cements, which have

TABLE 3.5 Comparison of rate constants of some C2S preparations [101]

80

Hydration of Portland cement

the advantage of low heat of hydration and improved durability. However the sluggish reactivity of β-C2S, as formed in Portland cement clinkers, means that the rate of strength development is too low for commercial applications. Much research has centred on the stabilization of reactive α′-C 2S, mostly through the addition of stabilizing elements [100]. Rapid cooling has also been a useful strategy [99,105], which has been developed for belite-rich clinkers at a pilot plant scale [106]. The effect of particle size on the β→γ transition is known to be important and can be described in terms of a critical particle size effect controlling nucleation of the transformation [107]. In recent years there has been renewed interest in reactive forms of β-C2S synthesized by chemical processes [108], spurred by the study of β-C2S formed by topotactic dehydration of hillebrandite [109]. However, the earliest chemical synthesis was reported by Roy and co-workers in the 1970s [110] and details of the early work are summarized by Ishida and Mitsuda [108] and Young [100]. Preparations from hillebrandite have surface areas of 2–8m2/g (depending on calcination temperature) and react completely within 28 days. Subsequently β-C2S was prepared by the Pechini process [111], a widely used chemical synthesis for many ceramic phases which utilizes a mixture of citric acid and ethylene glycol as the organic carrier phase. (The organic mixture can be replaced by poly(vinyl alcohol) [112] for better yield, but slightly lower phase purity.) In all these preparations â is the stable polymorph and β is difficult to obtain except when calcination takes place at high temperatures. This confirms the concept of a critical particle size for preventing the β→γ transition. The Pecchini process gives β-C2S with surface areas as high as 40m2/g, depending on the calcination temperature [113], and hence with proportionally faster reaction (Figure 3.15). The hydration kinetics are similar irrespective of the method of synthesis, and diffusion-controlled kinetics are only observed in the latter stages with samples that have the lowest critical surface area. The activation energy is comparable to that obtained for. conventionally sintered C2S [108]. The C-S-H produced has a C/S molar ratio approaching 2.0 (i.e. little crystalline

FIGURE 3.15 Effect of surface area on the time required for complete hydration. [Ref. 105, with permission.]

Ca(OH)2 is formed) but the silicate structure of CS-H does not appear to be changed significantly [114]. 3.4 Hydration of tricalcium aluminate Among the minerals present in Portland cement, C3A is the most reactive and is known to have a significant influence on the early hydration and rheology of Portland cement and concrete. The hydration behaviour of C3A, like that of other clinker minerals, depends on several factors such as temperature, w/s ratio, specific surface area of C3A, mixing procedure, size of the hydrating sample, and presence of admixtures. A review of the hydration of C3A and C4AF has been published by a RILEM Committee [115]. 3.4.1 Hydration in the absence of sulphate Breval [116] made an extensive study of the sequence of product formation in C3A hydration both in water vapour and in liquid, and confirmed the overall findings of numerous earlier studies. A gel-type material is apparently the precursor of the 81

E.M.Gartner et al.

hexagonal hydrates which grow first as poorly crystalline, thin, irregular flakes. With time, these flakes grow to better crystallized hexagonal plates with compositions C 2AH 8 and C 4AH 19. These crystals have a layered structure consisting of or OH– in positively-charged sheets with the interlayer region to balance the net positive charge of the layer, together with additional H2O. Some of the typical hydrate morphologies are shown in Figure 3.16 (see Section 3.7.4.2). The hexagonal hydrates are metastable and are converted to stable forms of cubic C3AH6.

(21) (22) C4AH19 loses part of its interlayer water to form C4AH13 at relative humidities below 88 per cent, so that the latter phase is detected when dried pastes are examined by X-ray diffraction. Also, in the presence of atmospheric CO2 part of the interlayer OH – is replaced by to form stable hemicarboaluminate [C4A(1/2OH–, 1/2 )H12].

FIGURE 3.16 SEM micrographs of the hydration products of C3A: (a) gel, consisting of irregular foils; (b) hexagonal flakes of AFm; (c) AFt in Portland cement paste. 82

Hydration of Portland cement

Figure 3.17 shows the heat-evolution profile of C3A hydrating in water. The time at which the conversion of hexagonal hydrates to the cubic phase takes place is strongly dependent on the conditions of the reaction, and is accelerated by the addition of C3AH6 nuclei. The high heat evolution, accompanying the formation of hexagonal hydrates, raises the temperature of the hydrating system appreciably, and can accelerate the conversion. Above 30°C the conversion of C4AH19 (or C4AH13) and C2AH8 is very rapid, but once C3AH6 has nucleated, crystal growth can readily occur even below 30°C. Direct formation of C3AH6 from C3A occurs at temperatures higher than 80°C, and can result in a strong matrix. However, conversion of hexagonal hydrates to C3AH6 leads to an increase in porosity and disruption of microstructure, which results in loss of paste strength.

FIGURE 3.17 Hydration of C3A: (a) heat evolution curve; (b) sequence of hydration.

3.4.2 Hydration in the presence of sulphate More relevant to the cement system is the hydration of C3A in the presence of gypsum, which is added to control setting. The ideal reactions are given in equation 23–25: (23)

(25) Ettringite is the first stable hydration product (equation 23) and forms as long as sulphate ions are available for reaction. Once the sulphate in solution drops below a critical value when all the gypsum is consumed, ettringite becomes unstable and further hydration yields the tetracalcium monosulphoaluminate hydrate (monosulphoaluminate) as the stable hydration product (equation 24). However, should the monosulphoaluminate become exhausted before the C3A is fully consumed then equation (25) will commence which describes a solid solution between and C4AH13 where x lies between 0.4 and 1.0. The establishment of a solid solution is due to the isostructural relationship between and and 2OH– occupy C4AH13 in which the ions interlayer positions in a random way. Other anions, for example, can also occupy interlayer positions, and since Fe3+ can also substitute for Al3+ in the layer, there is the possibility of wide compositional variations. Therefore, the compounds are collectively known as AFm phases. Similarly, the designation AFt is used to denote ettringite-type phases with similar (but more limited) variations in composition, that occur when portland cement hydrates. The reaction can be followed calorimetricaily as shown in Figure 3.18. The initial rapid hydrolysis of C3A and precipitation of ettringite is followed

83

E.M.Gartner et al.

FIGURE 3.18 Hydration of C3A in the presence of gypsum: (a) heat evolution curve; (b) classical mechanism of hydration proposed by Schwiete et al. [117].

by a period of slower reaction, as ettringite forms more slowly by a diffusion-controlled process. The renewed activity accompanies the transformation to monosulphoaluminate according to equation 24. The onset of the second peak depends on the amount of gypsum added. Figure 3.16 shows typical – morphologies of ettringite and C4ASH12. 3.4.3 Mechanisms of C 3 A hydration 3.4.3.1 In the absence of gypsum

When C3A is brought into contact with water, its high initial reactivity is followed by a period of slow reactivity. This is generally believed to be due to the initial hydration products, the hexagonal hydrates (C4AH13 and C2AH8), forming a protective barrier on the C 3A grain surface (Figure 3.17b). The conversion of these hydrates to cubic hydrates (C3AH6) disrupts the barrier and 84

the hydration proceeds rapidly again, accompanied by a large heat evolution. Supporting this mechanism are the findings that compounds, both organic and inorganic, that stabilize the hexagonal hydrates with respect to the cubic hydrates retard the C 3A hydration, whereas those that increase the rate of conversion, accelerate C3A hydration [118–121]. Corstanje et al. [122] further developed this concept of retardation by proposing the formation of amorphous Al(OH)3 on the reacting surface, due to localized concentration of Al3+in solutions close to the surface of the C3A grains. This is visualized as occurring when the hexagonal hydrates form an isolating layer to prevent rapid ion exchange with the bulk solution. As the hexagonal hydrates form, more Ca2+ is removed from the solution than is Al3+ and hence Al(OH)3, eventually precipitates. Periodic breakdown of the isolating layer, for example, by conversion to cubic hydrates, would allow the retarding Al(OH)3 layer to be attacked by Ca2+ and OH– ions in the bulk solution, and C3A would continue to hydrate. Skalny and co-workers [123,124] also attributed the retardation of initial C3A hydration to the formation of an Al-rich layer on the hydrating grain surface based on their observation of incongruent dissolution of C3A in water and dilute HCl. This layer, according to these authors, eventually re-arranges either to Al(OH)3 or to polynuclear aluminum complexes such as [Al8(OH)20]4+. Both these hypotheses, while ultimately predicting the formation of some kind of amorphous layer, approach the problem from quite different perspectives and were derived using data obtained under widely differing experimental conditions. The initial formation of an amorphous gel, first observed by Breval [116], is perhaps pertinent here. It can be considered to be either Al(OH) 3 or a co-precipitate of Al(OH) 3 and Ca(OH)2 which, through further reaction with Ca2+ and OH– ions from solution, forms the hexagonal hydrates. The exact role of an Al-rich layer in the overall hydration of C3A is not yet clear. Does it persist through the hydration process, or is it no longer capable of forming once C3AH6 nucleates and grows? To further understand the consequences

Hydration of Portland cement

of initial hydrolysis and early formation of precipitates, we need to investigate this part of the reaction in considerably greater detail. 3.4.3.2 In the presence of gypsum

Studies on the mechanism of C3A hydration in the presence of gypsum are of great practical value because of the influence of this reaction on the early reactions of Portland cement and concrete. Since ettringite is usually detected at the surface of the C3A, rather than on the gypsum particles, Schwiete and co-workers [117] proposed that it forms a continuous coating on the reacting C3A surface which impedes diffusion of , OH– and Ca2+ ions, slowing down the reaction and resulting in an induction period. The concept is shown schematically in Figure 3.18b. The more gypsum present in the system, the longer the induction period and the kinetics of ettringite formation indicate diffusion control [125,126]. When the liquid phase of the hydrating system becomes deficient in Ca2+ and ions, the protective coating is disrupted, which results in the renewed hydration of C3A to form AFm phases. Subsequently, the role of ettringite in the retardation of C3A hydration in the presence of gypsum was the subject of much discussion [115,122,127–131]. From their dissolution and electrokinetic data on C3A suspensions in acid media, Skalny and co-workers [123,124] concluded that the retarding effect of sulphate is due to its adsorption on reactive sites on the surface of C3A, a conclusion reached earlier by Feldman and Ramachandran [132]. Zeta-potential measurements show that the positive charge on the hydrating C 3 A particles (apparently due to adsorption of Ca 2+ on the Al-rich surface) is decreased by the adsorption of sulphate ions; chloride ions do not have any significant effect. The effect of sulphate ions in decreasing the number of active dissolution sites is probably achieved via a blocking effect of sulphate ions on coordination sites which would otherwise be occupied by hydrogen or hydroxyl ions, and which are known to catalyse the dissolution of alumina and other

oxides. Collepardi et al. [131] disagreed with the above conclusions because Na2SO4 does not have the same retarding effect as gypsum. Clearly, the ideal behavior as outlined in the equations 23–25 is more simple than the observed behaviour, which depends on the forms of sulphate available, the reaction of C3A and other factors. Evidence has been presented [130] for the formation of amorphous precipitates at the surface of C3A prior to the nucleation and growth of ettringite. Furthermore, monosulphoaluminate can co-exist with ettringite in the very early stages of hydration. Obviously equilibrium conditions are not established quickly and depend on the relative rates of solubility of C3A and the sulphate phase to establish the solution composition from which the initial products form. The composition of initial amorphous product will reflect the composition of the solution from which it precipitates and both will determine whether monosulphoaluminate will occur as a transient phase. Similarly, the presence of other sulphate phases such as hemihydrate , arcanite , or calci-uml langbeinite can lead to the formation of gypsum or syngenite and other intermediate products in hydrating cements. However, the sulphate ions eventually end up in AFt or AFm phases. Corstanje et al. [122] noted that at molar ratios where ettringite can be detected retardation was observed, while at molar ratios where ettringite could not form, gypsum tended to promote C 3A hydration. This they attributed to the ‘isolating ability’ of AFm, which does not promote Al(OH)3 precipitation to the same extent as do sulphate-free AFm phases, because Ca2+ ions are supplied by the gypsum in the system. Chatterji [133] disagrees and claims that because ettringite hinders the transport of sulphate ions from the bulk solution to the C3A surface, a retarding layer of sulphate-free AFm can form adjacent to the C3A surface. The above apparently contradictory observations and conclusions present a confusing situation. It must, however, be remembered that conditions of hydration were not the same in all studies; some were looking at dilute suspensions, while others were studying pastes of C3A or Portland cement. The same processes 85

E.M.Gartner et al.

may not occur in pastes and in suspension, and if they do, the times at which different processes occur are likely to be affected by the time at which critical ion concentrations are reached in solution. Ionic concentrations will also be affected by differing reactivities of different C3A preparations and, also by the presence of other cement minerals. Since AFm may form even when gypsum is still present in the system, [130,134] the reactivity of C3A compared to the rate of release of sulphate ions from gypsum crystals is a critical factor. Despite its importance in controlling the behavior of Portland cement pastes immediately after mixing, – the hydration of the C3A–CSH2 system has not been studied in the same depth as C3S. Nevertheless we anticipate many common elements between the behaviour of these two reactive phases. On that basis we propose the following scenario. 1. Initial rapid dissolution of C3A. 2. Almost immediate precipitation of an amorphous calcium aluminate hydrate gel, which acts as a protective coating. This gel probably incorporates a significant amount of anions, such as from the surrounding solution, as well as organic species if present. 3. Nucleation of crystalline products from the amorphous layer: AFm or AFt depending on the particular circumstances. Minor components, such as alkalis, chlorides and admixtures, may also enter the gel and modify the crystal growth by random substitution in the lattice of the hydrates. 3.4.3.3 C 3 A hydration under alkaline conditions

In cements, C3A and sulphate ions react in the presence of saturated Ca(OH)2 solutions derived from C3S hydration, and in many cases in the presence of alkali sulphates will raise the pH still further. When both Ca(OH)2 and gypsum are present the hydration of C3A is more retarded than by gypsum alone [131,132,135]. The result is slower formation of ettringite, and the conversion to AFm can be delayed for several days. The slower reaction has been attributed to the development of a 86

more effective diffusion barrier: colloidal-sized ettringite [127] or C4AHx [138]. In such systems a gel-like phase precedes the growth of crystalline ettringite [130]. Alkali compounds in general accelerate the hydration of C3A. Spierings and Stein [136] observed that in a sulphate-free environment <0.4M NaOH decreased early hydration (ascribed to decreased solubility of Ca2+), while >0.4M NaOH accelerated hydration. This latter effect was attributed to the OH– catalysed hydrolysis of the Al–O network and a reduced stability of AFm with respect to C3AH6. However, in cements, alkalis are usually present as sulphates. Tang and Gartner [137] blended a low alkali clinker with various combinations of gypsum and alkali sulphates which are found in most high alkali clinkers. The reaction of C3A was enhanced – by aphthitalite K3NS4, but reduced by syngenite or calcium langbeinite. Alkalis are also incorporated into the C 3A structure during the clinkering process. The presence of varying amounts of Na2O in C3A can result in as many as four different crystalline modifications: cubic, orthorhombic, tetragonal and monoclinic. There is general agreement that the hydraulic reactivity of C3A markedly decreases as the Na 2O content in the crystal increases; [136,138–142] for example, cubic C3A (pure phase) is more reactive than the monoclinic polymorph (5.7 per cent Na2O), as shown in, Figure 3.19) [139]. For Na2O-doped C3A, the time of conversion of hexagonal hydrates to cubic hydrates during hydration, as observed by calorimetry, increases considerably and the size of the heat-evolution peak decreases markedly [139]. Other ‘impurities’, such as Cr2O3, TiO2, MgO, Fe2O3 and SiO2, have a similar effect [140]. The reactivity of a C 3A polymorph has been found to depend on the extent to which the lattice holes are occupied by the ‘impurities’ and it decreases with increasing occupancy [140,141]. 3.4.4 Hydration in the presence of other species C3A hydration in Portland cement should not be expected to be independent of the presence of

Hydration of Portland cement

phases. Calcium carboaluminate hydrate forms from C3A–CaCO3 mixtures [148] and almost certainly forms during the hydration of Portland cements containing ground limestone, although it is not easily detected. 3.5 Hydration of the ferrite phase

FIGURE 3.19 Hydration of Na 2 O-doped C 3 A polymorphs [138].

either C3S or C2S, and indeed an interaction exists. Regourd and co-workers [142,143] found that when C3A was hydrated in the presence of C3S, monoand trisilicoaluminates (C 3 A·CS·H 12 and C3A·3CS·H31 respectively) were formed around C3A by diffusion of silicate ions from C3S through the solution. The monosilicoaluminate first formed outside the original C3A grain boundary; after 7 days, poorly, crystalline trisilicoaluminate was detected. These silicoaluminates were unstable in the presence of sulphate ions and did not appear in samples with 20 per cent gypsum. The interaction of C3S was greater with cubic C 3A than with tetragonal C3A. The addition of other forms of silica, for example, ground quartz accelerates C 3A hydration [144,145], although natural pozzolans retard [146]. Effects at later ages may differ from those observed initially. The addition of chloride accelerates the formation of ettringite in C3A-gypsum mixtures [147]. Only when gypsum is exhausted does the chloride AFm phase (calcium chloroaluminate hydrate) form. (This phase also will form as chloridebased deicing salts penetrate concrete.) Other species present in chemical admixtures (e.g. nitrate and organic carboxylates) may form modified AFm

C4AF is not a stoichiometric compound, but part of a solid-solution series called the ferrite phase. It has the general formula Ca4Fe(2–x)AlxO10, where x varies from 0 to 1.4, C4AF is just one point in the series where x=1.0. [It can be considered as a limited solid solution between C2F and hypothetical C2A.] Many cements have a ferrite phase with a composition close to C4AF and so this is the most common composition studied. However, the reactivity of the ferrite phase decreases as the iron content increases [115,149]. The course of hydration of the ferrite phase is very similar to that of C3A [149–157]. In the absence of gypsum, the metastable iron-substituted AFm phases C4(A,F)H 19 and C2(A,F)H8 form, which convert to C3(A,F)H6. These crystalline phases have a higher A/F ratio than does the unhydrated material. Thus, to maintain stoichiometry an iron-rich hydroxide must form. This phase is amorphous to X-rays, but has been detected by SEM [152] and Mossbauer spectroscopy [154]. Thus the reactions can be written as: (26a) (26b) Although the exact composition of the crystalline phases appears to be highly variable, a complete disproportionation of iron is unlikely. However, C 3 (A,F)H 6 is unstable at high temperatures, decomposing to C3AH6, and α-Fe2O3 (haematite) [151]. When gypsum is present the hydration sequence for the ferrite phase is the same as for C3A. Ironsubstituted AFt first forms and later converts to AFm when the gypsum is exhausted. An amorphous phase, presumably aluminum-substituted Fe(OH)3 is also formed. Disproportionation of iron between the crystalline and amorphous phases is observed 87

E.M.Gartner et al.

here too, and is attributed to the low mobility of iron. The formation of iron-rich pseudomorphs of the original C4AF grain has been observed both in pure systems [158], and in cement [159]. The acceleration of C 4AF hydration with tertiary alkanolaminines is attributed [160] to an increase in the mobility of iron, through the formation of Fe3+-amino complexes. Brown [161] has concluded that C2AH8 or alumina gel precedes the formation of AFt in the C4AF-gypsum system and that early AFt is essentially iron-free. Ramachandran and Beaudoin [157] found that the sequence of reactions and interconversions of hydration products at low w/s ratios and high temperatures, were different from those occurring at the normal w/s ratio and temperature. The conversion of ettringite to AFm phases could occur even in the presence of gypsum. Also, at very low w/s ratios and higher temperatures, AFm phases did not necessarily result from reactions involving ettringite. It is generally assumed that C4AF is less reactive than C3A, but this may not necessarily be true in actual cements. It must be remembered that impure phases are present in Portland cement. The reactivity of C3A depends on alkali content and it may contain up to 10 per cent Fe2O3 (corresponding to C9A8/3F1/3). The ferrite phase in Portland cement contains SiO2, MgO and alkalis. The effect of these substitutions might be to lower the reactivity of C3A and increase that of the ferrite phase [154], thereby lessening the difference between their reaction rates. C4AF is much more strongly retarded by the addition of gypsum or a combination of gypsum and lime than is C3A (Figure 3.20) [150,154], although the reasons for it are not clear. The persistence of iron-rich amorphous coatings that do not transform to crystalline products may be the key to retardation, although the formation due to stable coatings of AFt have been proposed [154]. 3.5.1 Hydration in the presence of C3A Since their hydration is so similar, C3A and the ferrite phase must compete for gypsum during the hydration of Portland cement. This interaction was studied by

88

FIGURE 3.20 Heat evolution of (a) C3A and (b) C4AF during hydration in water, in saturated solution of gypsum, and in saturated lime-gypsum solutions [149].

Jawed et al. [149], who added small quantities of C3A to mixtures of C4AF and gypsum. The conversion of ettringite to AFm occurred sooner, depending on the amount of C3A added (Figure 3.21). Ettringite is formed more rapidly indicating that C3A competes more efficiently for sulphate ions. Once depletion of gypsum occurs, which happens earlier in the presence of C 3 A, C 4 AF reacts to form monosulphoaluminate. 3.5.2 Hydration of calcium ferrite Calcium ferrite (Ca2Fe2O5 or C2F) can occasionally occur in Type V cements. The hydration of C2F has been shown [156] to produce C4FH13 (which initially forms as C4FH19) and amorphous Fe(OH)3. The C4FH13 converts to C3FH6 within a day, but the latter rapidly breaks down to CH and haematite. Teoreanu et al. [154] did not observe either C4FH19 or C3FH6 in their study of C2F hydration. However, it has been pointed out that these hydrates give the same Mossbauer spectra as Fe(OH)3, and that Xray diffraction (XRD) is not sensitive unless a good monochromator is used to filter out iron fluorescence [161]. Rogers and Aldridge [101] also studied the hydration of ‘C3F’ and ‘C4F’ which are actually intimate mixtures of C2F and C. As expected, they behave similarly. CF also hydrates in the same way, but much more slowly.

Hydration of Portland cement

FIGURE 3.21 Hydration of C4AF in the presence of C3A with either gypsum or hemihydrate [149] G=gypsum, H=hemihydrate, E=AFt, M=AFm.

3.6 Hydration of Portland cement

summarized in Table 3.6. The hydration reactions proceed until either a lack of reactants (cement components, water), or a lack of space to deposit the hydration products, causes the reactions to cease. The hydration of a Portland cement is much more complicated than the hydration of individual ‘pure’ clinker phases, because the reactions of the different phases proceed simultaneously at differing rates and thus, can influence each other in complex ways. However, the general principles are the same: the dissolution of anhydrous phases leads to the precipitation of much less soluble products, typically colloidal and micro-crystalline hydrates, that form the hardened paste. An oversimplified but convenient way of following the progress of cement hydration reactions is by the use of heat evolution curves (Figure 3.22), such as are used in studying the pure components, they are useful in correlating the known experimental data to a single parameter, heat evolution. There are at least two limitations to this approach: (1) the rate of heat evolution is neither simply proportional to the degree of hydration of any one phase (or even to the average of all of the phases, nor to the development of physical properties; (2) isothermal hydration studies do not properly represent hydration processes in most engineering applications where concrete develops mechanical properties in semi-adiabatic rather than psuedo-isothermal conditions.

3.6.1 The hydration process

3.6.1.1 The first minutes of hydration

The hydration of Portland cement is a sequence of overlapping chemical reactions between clinker components, calcium sulphate and water, leading to setting and hardening. In the major application, that is in concretes (or mortars, etc.) the setting process is the consequence of a change from a concentrated suspension of flocculated particles to a viscoelastic skeletal solid capable of supporting an applied stress, and can be monitored by rheological measurements [162]. The continued development of the solid skeleton is called hardening—a physico-chemical process, which leads to the development of ultimate mechanical properties. The principal stages of hydration are

As soon as the cement contacts water, a series of rapid reactions begins which involve mainly the clinker interstitial phases (i.e. the aluminates and aluminoferrites, alkali sulphates, and free lime), plus the calcium sulphates (gypsum, hemihydrate and/ or anhydrite) which have been interground with the cement. The aluminate phases (including the aluminoferrites in most cases) react very rapidly and exothermically, giving a flush of calcium and aluminate ions into solution. The only component which is not usually observed going through solution is the iron, which probably remains mostly undissolved in the form of a hydrated oxide coating on the surface of the aluminoferrite phase [161]. 89

E.M.Gartner et al. TABLE 3.6 The four principal hydration stages of Portland cement concretes

However, the high aluminate concentrations do not last long. In a matter of seconds there is precipitation of an aluminate hydrate layer over the surfaces of the cement particles, this layer being generally referred to as being ‘ettringite’ or ‘AFt.’, although it is often poorly crystalline and difficult to detect by XRD at very early ages. As expected 90

for an AFt phase, it contains a lot of sulphate, which comes from the rapidly soluble sources such as alkali sulphates, hemihydrate and gypsum. Free lime in the clinker usually dissolves rapidly and exothermically, and, if present in sufficient quantity, can lead to portlandite supersaturation almost immediately; however, some may also be

Hydration of Portland cement

FIGURE 3.22 Schematic representation of heat evolution during hydration of a cement.

consumed by reaction with the aluminate phases. Reactive alite particles also release CH rapidly into solution, so it is not always easy to separate the influence of alite and free lime at early ages. Alite usually contributes much less to the initial heat release than do the aluminates, but the amount coming from free lime can vary widely. Dissolution of calcium sulphates is only weakly exothermic and relatively little dissolves in the first few minutes; on the other hand, dissolution of alkali sulphates is very rapid but endothermic. The solution usually reaches a fairly constant alkali concentration within less than a minute, when all of the exposed crystals of alkali sulphate have dissolved completely. Alkali metal ions are not precipitated in any of the early reactions except for the , which only occurs formation of syngenite at high potassium concentrations. Rapid precipitation of sulphate always occurs, principally as aluminate hydrates, but often also as gypsum, and/or syngenite. To balance the dissolved cations (mostly alkalis) this missing sulphate is replaced by hydroxide ions: thus, the pH increases very rapidly in the first minute, and soon becomes dependent almost entirely on the total alkali metal concentration in the liquid phase, (denoted by [R+]) the anion concentrations in most typical cases

being determined by a state of near equilibrium with gypsum and ‘supersaturated’ CH [163]. The initial dissolution reactions can be represented chemically by the following equations, in which all of the ions shown are intended to represent fully solvated aqueous species: (27) (28) (29) (30) (31) We have chosen to represent the initial coating that forms on alite surfaces (and probably also on belite surfaces) as ‘C-S-H(m)’ by analogy with the pure C3S system, and ignore the fact that it probably forms via a rapid solution-precipitation reaction and so may incorporate other ions; similarly, we represent the iron-rich coating that forms on the aluminoferrite phase as ‘FH3’. However, the actual compositions and structures of these superficial phases are unknown, and their spatial extent may be influenced by the adjacent phases. 91

E.M.Gartner et al.

These dissolution reactions are followed immediately by rapid precipitation of aluminate hydrates: (32) (33) It is by no means certain that all of the aluminate precipitates as ettringite as shown in equation (32); AFt phases in which some sulphate is replaced by silicate or hydroxide have been reported [164,165]. In cases of calcium sulphate deficiency AFm phases such as the hydroxyaluminate (X=OH – ) or monosulphoaluminate (2X=SO 42– ) (or a solid solution of both) can form instead, as shown in equation (33). There may also be a little iron substitution in these phases. As stated earlier, the AFt phase that forms early on often shows very weak XRD peaks and is rather difficult to characterize morphologically; it is sometimes claimed to be amorphous. In addition to the hydrated aluminate phases, which precipitate in significant amounts within the first minutes (of the order of 1–5 per cent by mass of cement), there is often substantial precipitation of hydrated sulphate phases (gypsum or syngenite, or both,) from cements that contain high levels of ‘metastable’ sulphate sources (principally calcium sulphate hemihydrate and calcium langbeinite) [136]. Precipitation of these phases occurs at about the same time as that of the aluminate hydrates, and all of these precipitation reactions can lead to observable changes in workability if they continue after the concrete has left the mixer. 3.6.1.2 The induction and acceleration periods

The hydration of C3S has often been used as a model for the hydration of Portland cement. This practice is based on the fact that Portland cement clinkers typically comprise about 75–80 per cent alite and belite (impure C3S and C2S respectively), and that C-S-H produced in the early stages of hydration is predominantly due to alite hydration. 92

Using this analogy, it is common to refer to the period between mixing and set as the ‘induction’ period, although it is not a period of complete inactivity. In fact, most Portland cements show a significant level of heat evolution throughout this period, but it is difficult to determine how much of this heat comes from the silicate phases and how much from the aluminates. The mechanism necessary to explain this period of slow activity in portland cements need not be very different from that of the pure systems. In most cases, the order of events appears to be similar to a simple combination of those observed in the two – separate systems: C3S–water and C3A–C S –CH– water. Both systems show an induction period after an initial fast reaction, although, in the aluminatesulphate system, the length of this induction period depends very strongly on the ratio of sulphate to C3A [146], and is also sensitive to the presence of CH [128,135,166]. It is quite reasonable to suppose that the two systems behave fairly independently from a chemical standpoint. Microanalyses suggest that initially C-S-H is admixed with AFt, AFm and CH at a nanometer level, but that C-S-H itself has relatively low degrees of aluminate or sulphate contamination [8]. Thus, the behaviour of the two systems probably become inter-related physically [167] and this could have a big influence on the actual rates observed in the combined systems, even if the chemical mechanisms are not fundamentally changed. Bearing this in mind, there is fairly good evidence that the acceleration period observed in pure C3S hydration occurs very similarly in most normal Portland cements, and it is usually the formation of C-S-H during this stage that is responsible for the set and early hardening of concrete. According to Bezjak and Jelenic [67], the presence of C3A and soluble sulphates significantly extends the acceleration period, but, once the diffusion stage begins, the rates become slower than in the pure C3S system. However, the hydration of the aluminate phases is probably even more strongly modified from – that observed in the C3A–CS –CH–water system. In most Portland cements, there is sufficient sulphate available, in principle, to prolong the C3A induction

Hydration of Portland cement

period well beyond the first day; thus, very little C3A hydration is occurring at the time at which the set occurs, except in specially modified Portland cements. Kanare and Gartner [168] suggested that it was necessary to assume that calcium sulphate was significantly adsorbed (or, possibly, physically entrapped) by the C-S-H produced during rapid C3S hydration, in order to explain why the re-acceleration of C3A hydration usually occurs much earlier than expected in portland cements. Electron microscopy studies have not always aided in the comprehension of the hydration mechanism, due to the complexity of the microstructure and the difficulty in analysing small volumes of hydrate phases. Double et al. [60], using an analogy with ‘silicate gardens,’ proposed an ‘osmotic pump’ caused by the formation a semipermeable silicate hydrate envelope, separating the bulk liquid from the anhydrous surface, after mixing cement with water. This envelope eventually ruptures under an increasing osmotic pressure forming hollow tubules. Later studies [169] showed these tubules to be AFt. Osmotic mechanisms are not observed for pure C3S and the presence of alumina (and perhaps sulphate or other species) may be necessary for the formation of the proposed semi-permeable membrane. This mechanism has not stood the test of time. However, frequent observations of ‘hollow shell’ hydration [170] (see Figure 3.23) in most portland cement pastes seems to be consistent with a modification of the hydration processes, at least for some of the particles in the cement. The excellent early work of Pratt and co-workers [171,172] was followed by a more detailed analysis of the sequence of events leading to the formation of hollow shells [158,167]. Since hollow shells are not observed in pastes of pure C 3S they probably originate from multi-component grains due to concomitant precipitation of AFt and C-S-H. Dalgleish et al. [172] attempted to correlate the main heat evolution peak of cement hydration to a new point in the process of hardening, which they termed the ‘cohesion point’, at which time the bonds between particles supposedly become strong enough to fracture, under stress, the thickening shells and thus expose unhydrated particles. In our

FIGURE 3.23 Back-scattered electron images of cement paste, ×1000 (a) showing phenograins (p). (Micrograph courtesy of Dr. P.Stutzman.) (b) showing hollow hydration shells and calcium hydroxide (white) adjacent to an aggregate particle. (Micrograph courtesy of Dr. A.Crumbie and Dr. C.Famy.)

view, the above correlation is questionable because the top of the heat evolution peak represents the onset of diffusion control. Cements with coarser particles have a tendency to form a different shell structure from that formed by finer cements, and this may have an effect on the early strength development and Young’s modulus. 3.6.1.3 Overall kinetics of hydration

Although most of the cement reacts with water within the first few days, the rates at which the individual 93

E.M.Gartner et al.

clinker compounds react are different. Ranking them in terms of their average rate of disappearance over the first few days or weeks of hydration, their overall reactivities usually decrease in the order: C 3 A>C 3 S>C 4 AF>C 2 S. However, the effective reactivities of two cements of similar bulk chemical and mineralogical compositions may differ significantly, because the rates at which the individual components react with water depend on the processing history, primarily on the clinkerburning rate and temperature, and the cooling rate [173]. These factors affect the distribution of mineral phases in clinker; their crystal size, concentration of crystal imperfections particularly at phase boundaries crystallinity, and polymorphic form of some of the clinker components. Clinker composition and reactivity also depend on the mineralogy of the cement raw mix components. Additional differences in effective cement reactivity result from comminution—principally differences in psd. Because of the different hardness and fracture toughness of the various clinker components, differential grinding leads to uneven distribution of individual components amongst the various size fractions. This effect is particularly marked with calcium and alkali sulphates, which are much softer than the clinker phases, and tend to be found preferentially in the finest fractions of the cement. However, such particles can have a tendency to self-agglomerate, which means that they may behave as much coarser particles unless they are well dispersed in the cement. The distribution of the calcium sulphate phases in the cement, as well as their solubility rates, can have a very significant influence on the course of the rapid early reactions of the aluminate phases [136,174]. Tang [174] found that a cement with poorly distributed gypsum particles (produced by interblending) is less well controlled than one of identical mineralogical composition and apparently similar psd, but with well-distributed gypsum particles (produced by inter-grinding). The aluminate phases in the latter cement only reacted to about the half extent of those in the former during the first two minutes of hydration. This is consistent with the concept that formation of a thin protective hydrate layer on the aluminate phases 94

requires a very high initial flux of calcium and sulphate ions close to the surfaces of the aluminate particles [136]. These observations are just one facet of the general observation that the rate of hydration of clinker minerals depends on the composition of the liquid phase surrounding them, especially if dissolved species (either those originating from the other cement phases, or admixtures deliberately added to modify the hydration reactions) are involved in the reaction in question. There are also cases in which the presence of fine, but essentially insoluble, particles can have strong influence, if they are well distributed. One such is the observation that fine calcite particles generally accelerate early cement hydration [175– 177], probably by acting as potential nucleation sites for the reaction products. A similar example is the increased rate of reaction of C2S in the presence of C3S [178,179], suggesting that the C-SH produced by the hydration of the C3S can serve as a local growth site for the products of C2S hydration. C3S controls the solution composition even when only small amounts are blended with C2S [93]. Differences in typical rates of hydration of pure clinker minerals and those of the same compounds in cement are illustrated in Figure 3.24. The differences are not very great, the two principal effects in the cement paste being, as explained above, the shorter than expected induction period for the C3A, relative to the pure C3A-gypsum system, and the faster than expected hydration of C2S. The temperature of hydration is another important kinetic factor. Generally, increased temperature accelerates the early rate of cement hydration (Figure 3.25), although the degrees of hydration and strength development at later ages are often decreased. In mass concrete, where near-adiabatic conditions are maintained, the evolution of heat of hydration will raise the temperature to well above ambient. The generation of high internal temperatures occurs primarily over the first few days when the concrete is still relatively plastic, and will not cause immediate damage, but potential effects on strength development are significant [180,181], illustrated in Figure 3.26. On subsequent cooling when the

Hydration of Portland cement

FIGURE 3.26 Compressive strength development of concrete cured under standardized laboratory conditions, and under a temperature regime mimicking internal concrete temperatures in the field [182].

controlled by the addition of mineral admixtures such as fly ash or blast furnace slag [183]. 3.6.1.4 Energetics of hydration FIGURE 3.24 Rate of hydration of the cement compounds: (a) in pastes made with pure compounds; (b) in a Portland cement paste [180].

FIGURE 3.25 Effect of curing temperature on compressive strength development.

concrete has gained rigidity, thermal stresses resulting from temperature gradients may well lead to tensile cracking. Temperature rise can be controlled by lowering the total heat of hydration of a cement through a reduction in C3S and C3A contents, because these compounds have the highest heats of hydration and the highest rates of heat evolution. More commonly, heat evolution is

Hydration of cement involves mainly exothermic reactions, and most of them also have positive ‘activation energies.’ Thus, if cement hydration is conducted under adiabatic or semi-adiabatic conditions, it will self-accelerate due to the increase in internal temperature, regardless of the specific chemistry. The initial burst of heat produced mainly by aluminate and free-lime hydration is usually enough to raise concrete temperatures in the mixer by about 2°C. Once alite hydration enters its acceleration period, the temperature of the concrete rises rapidly, typically by 20–40°C. Selected heats of hydration for the main clinker components are given in Table 3.7 in joules per gram (rather than joules per mole) as a more practical measure for the cement technologist. The enthalpy of hydration, ∆H, represents the differences in the residual energy in the system after breaking the old and forming the new chemical bonds. Using these ∆H values, together with the mineralogical and chemical compositions of the cement and the degree of hydration of individual clinker minerals one could, in principle, estimate the heat of hydration of a cement at any given time. However, ∆H for many hydration reactions is still not known with certainty, and it 95

E.M.Gartner et al. TABLE 3.7 Heats of hydration of the cement compounds

a

Determined from regression analyses of heats of hydration of ground clinker or cement.

is experimentally difficult to determine the exact clinker composition and the actual degrees of hydration. The ∆H values given in Table 3.7 for cement and clinker have been determined by multiple regression analyses of heat of hydration conducted on a series of cements and clinkers hydrated for 1 year, and thus do not necessarily represent complete hydration. 3.6.2 Setting of cement paste One of the first physical consequences of the chemical processes between cement components and water (described earlier) is the transformation of a workable plastic cement paste (fresh paste) into a rigid, brittle material (hardened paste). This solidification process is the result of the formation of interlocking hydration products and is commonly described as setting. In general terms, however, setting depends on the relative rate at which the ionic species released from the cement components into the liquid phase react to form initial hydration products, which can be either colloidal or crystalline in nature. 3.6.2.1 Forms of setting

Setting encompasses ‘initial’ and ‘final’ set, as specified by various national standards (e.g. BS12, ASTM C150, DIN 1164) using penetration methods. These terms do not have a well-defined physical meaning. However, recent studies have shown that initial set can be determined rheologically from the 96

rate of change of yield stress [162] or ultrasonically from the rate of propagation of shear waves [184]. These measurements emphasize that setting is not dependent on hydration per se, but is a microstructural parameter representing the onset of percolation of a network of agglomerated solid particles. Thus, setting time depends on w/c and this is a major reason for poor correlation between the setting of cement paste and the setting of concrete. Most investigators have concluded that normal setting is controlled by the hydration of alite. Initial set under normalized conditions is reported [182] to correlate with the end of the induction period, although this time is not always well defined. Final set is said to occur about mid-way through the acceleration period. 3.6.2.2 Abnormal setting

Abnormal setting behaviour can, in most instances, be traced to chemical reactions involving the aluminates (C3A and C4AF) and sulphate phases (calcium and alkali sulphates). In some cases, however, highly reactive or unreactive alites (a consequence of burning/cooling kiln conditions or other factors) may be the cause. The most common types of abnormal setting are discussed briefly below. Chemical admixtures can result in abnormal setting behaviour in some cases, because they may change the delicate balance between the aluminate and sulphate precipitation reactions in an otherwise normally setting cement.

Hydration of Portland cement

False set (also referred to in the literature as early set or premature stiffening), is defined as a rapid increase in paste rigidity without excessive heat evolution, which can be dispelled by immediate further mixing to regain plasticity. False set generally results from rapid crystallization of gypsum from the paste liquid phase supersaturated in sulphates by excessive dissolution of calcium sulphate hemihydrate (plaster of Paris) or soluble anhydrite present in cement. Hemihydrate usually forms as a result of the dehydration of gypsum during grinding or during storage, especially if the cement is hot. Occasionally, the formation of ettringite or syngenite will cause false set. Flash set (also called quick or grab set) is defined by a strong heat evolution and a very severe stiffening, which cannot be dispelled by remixing. It is the result of an excessive reaction of C3A with water in the earliest stage of hydration to form AFm phases, due to an excessive rate of dissolution of the aluminate phases relative to the rate at which sulphate and calcium ions can be transported to the reacting aluminate surfaces. It has been attributed to an inadequate content of rapidly soluble sulphate; however, calcium hydroxide is also required, and cements with very unreactive alites or low free-lime levels can thus be susceptible. A special case of abnormal setting is pack set (also called warehouse or silo set) is characterized by a decrease in the ease of flowability of the unhydrated cement powder. Although this type of abnormal set happens during cement storage rather than after mixing cement with water, pack set can influence the subsequent hydration reactions and thus, is of interest to this discussion. Pack set is related to the degree of gypsum dehydration during cement grinding and to silo temperature and humidity, which, in turn, are functions of mill temperature and the amount of water added to the cement during grinding. It is most commonly due to strongly attractive surface forces (probably multipolar electrostatic interactions) between cement particles, and these can be reduced greatly by certain small organic molecules, such as glycols, acetic acid, etc., that can adsorb on the particle surfaces and make them more mutually

repulsive. Such additives are referred to, in this application, as ‘pack-set inhibitors;’ but they have a similar dry-dispersant function during grinding and so are normally added as grinding aids to improve powder flow and dispersion in the mill and separator. Gypsum in the warmer parts of the silo may dehydrate, (to give hemihydrate, or to react with potassium sulphate to form syngenite), releasing water vapour that can superficially hydrate other clinker phases such as the aluminates and free lime. Condensation of water near the colder silo walls may also cause formation of wall coatings via cement hydration. These silo reactions generally lead to a cement with a reduced rate of aluminate phase dissolution and, at the same time, an increased the rate of solubility of calcium sulphate, (because of the higher solubility rate of hemihydrate compared to gypsum.) This can lead to false set in concrete, and also to poor early-strength development due to slow hydration of the clinker phases. 3.6.2.3 Mechanisms of setting

In the classical literature [185], the setting or stiffening of cement paste is recognized to be caused by the increasing volume of hydration products, leading to a decrease in the distance between individual particles until plastic flow is restricted by cohesive forces. Although most researchers accept the importance of alite hydration, some believe that setting is controlled by recrystallization of ettringite. Rhebinder and co-workers [186,187] proposed setting to be the result of the formation of a coagulated thixotropic structure from the dispersion of unhydrated cement components by the initial colloidal hydration products (setting), with products from the aluminate phases being the dominant factor. Locher [130,188] concluded that set control is not the result of a chemical reaction, but rather of the form and size of crystalline hydrates produced, which depends on the relative availability of aluminate, sulphate, and calcium ions in the solution. When sulphate is present, minute ettringite particles deposit on the cement grains. Because of

97

E.M.Gartner et al.

their fine structure, plasticity of the system is maintained and normal set is the consequence of recrystallization, during the induction period, of finegrained ettringite into large needle-like particles, which produce a rigid structure by bridging the adjacent cement particles. Only a small portion of the C3A, dissolved in the first minutes of hydration, participates in these reactions. Normal setting is achieved when the amount and rate of availability of aluminates and sulphates enables complete coverage of the available surface by fine-grained ettringite. Inadequate sulphate causes the formation of large tabular crystals of AFm in addition to ettringite, leading to quick set. An excess of sulphate in the solution may lead to false or quick set, depending on the amount of secondary gypsum formed. Other authors [123,189,190] have shown, however, that retardation of C3A hydration by gypsum does not necessarily involve the formation of ettringite. Also, as noted earlier, it has been demonstrated that the presence of soluble calcium sulphates can often have a strong influence on the amounts of C3A and C4AF that hydrate in the first few minutes [136]. The amounts of these two phases consumed in the first two minutes can be reduced by at least a factor of three on going from a pure clinker with no added sulphate, to a cement with a suitable combination of soluble sulphates. The recrystallization of ettringite is probably not the usual cause of abnormal set. However, it is often observed in cement pastes about the time that the free sulphate disappears, which is after set [167]. This transformation can be attributed to the increased solubility of aluminate ions when sulphate concentrations are reduced, allowing some of the ettringite to re-dissolve and either recrystallize locally (probably mainly outside the hydration shells which surround cement particles at that time), or contribute sulphate ions for continued formation of AFm phases (probably mainly inside the shells) [167]. We concur with the view that normal set is the consequence of the formation of C-S-H resulting from C3S hydration. The formation of fine-grained ettringite from a solution of proper aluminatesulphatecalcium proportions is desirable to maintain a fluid 98

paste through the induction period until the formation of C-S-H causes the system’s mobility to decrease. All forms of abnormal set result from imbalance of ionic species in the liquid phase, and are the result of either improper proportions of C3A and sulphate in the cement, or differing reactivities (solubility) of sources supplying the aluminate and sulphate ions. 3.6.3 Liquid phase composition It is clear from the above discussion that the dissolution of ionic species and composition of the liquid phase play an important role in cement setting. The absolute amount of C3A in the clinker is of less importance than the reactivity of C3A and the solubility of calcium sulphate phase present in cement (gypsum hemihydrate soluble anhydrite). The concentrations of sulphate ions in the pore solution are very much influenced by the presence of soluble alkalis in the cement clinker [191] (see Figure 3.27). Gartner et al. [163] showed that the composition of the aqueous phases extracted from different cement pastes could be explained roughly on the basis of established solubility. The solutions were initially supersaturated with respect to gypsum and undersaturated with respect to CH, but fell close to gypsum saturation levels after the first few minutes, while slowly becoming supersaturated by a factor of 2–3 with respect to CH after a slightly longer period. High-potassium cements gave solutions close to syngenite saturation most of

FIGURE 3.27 Composition of pore solution in a highalkali Portland cement paste (w/s=0.5).

Hydration of Portland cement

the time, so that the liquid phase at early ages can be regarded as a perturbed portlandite-gypsum (– syngenite) system. Their data for early ages were consistent with previous results obtained over a wider range of hydration times, as summarized in Figure 3.27 [129,192]. The disappearance of sulphate ions from solution